Enhancing Egg Grading Precision through AI and Computer Vision-Powered Morphometric Analysis

Enhancing Egg Grading Precision through AI and Computer Vision-Powered Morphometric Analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Enhancing Egg Grading Precision through AI and Computer Vision-Powered Morphometric Analysis Henna Hamadani, Ambreen Hamadani, Pakcha Hannah Boje, Amelia Moyon, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6016705/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Egg size determination is an important activity in the poultry industry. Traditional methods of size assessment are labour-intensive, error-prone, and time-consuming. Machine learning is proving to be a major game changer in all sectors and this has the potential to upgrade and automate egg grading as well. Considering all this, this research was undertaken to evaluate the potential of AI-driven computer vision approaches for the extraction of egg dimensions from 2D images. The images were annotated and saved as an OBB dataset and were appropriately preprocessed. Yolo11 OBB models were evaluated for their ability to detect eggs and yolo11x-obb model was found to be optimal for this study. The Mean Absolute Errors (MAE) of the two final predictive models for egg size were 0.28 for length and 0.19 for breadth, with Pearson Correlations of 0.81 for length and 0.85 for breadth for Model 1 and MAEs of 0.29 for length and 0.26 for breadth, with Pearson Correlations of 0.88 for length and 0.81 for breadth for Model 2. Science4ImpactStatement : This foundational research concludes that Machine Learning methods can predict egg dimensions with good accuracy and could potentially be used for poultry breeding, automatic egg grading, packaging, processing as well as marketing. Computer Vision Egg Dimensions Artificial Intelligence YOLO Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Egg size measurement is one of the most important aspects both in hatchery management and in the poultry industry. This is an important aspect that influences pricing, packaging, and consumer preference and the evaluation time between laying and consumption is a crucial factor [ 1 ]. Traditional methods of egg size assessment, often involve manual measurement or mechanical devices. These methods are labour-intensive, prone to human error and time-consuming. They may also cause production losses during handling [ 2 ] and due to these reasons are also expensive in the long run [ 3 ]. In addition, egg weight, as well as egg quality are important indicators of hens laying performance. They represent factors which affect the health and overall well-being of laying hens [ 4 ]. The advent of computer vision and machine learning technologies offers a promising avenue for automating this process, enhancing both efficiency and accuracy. Machine learning has been successfully applied in animal sciences for tasks such as livestock behaviour recognition [ 5 ], disease detection [ 6 ], and animal identification [ 7 ], making breeding decisions [ 8 ], building decision support systems [ 9 ], making futuristic predictions [ 10 ] and for farm automation [ 11 ]. Recent studies have demonstrated the growing potential of computer vision systems in egg grading and defect detection. For instance, [ 12 ] developed a two-stage model combining real-time multitask detection and random forest networks to predict egg categories and weights, achieving high accuracy in sorting and weighing eggs. (Ab Nasir et al., 2018). Similarly, [ 13 ] investigated an automated egg grading system using shape parameters, finding that shape-based grading yielded better recognition results compared to weight-based methods. However, existing approaches often face challenges such as limited generalizability across different egg types, sensitivity to environmental variations, and computational inefficiencies. [ 14 ] presented image processing and machine learning techniques for predicting chicken egg weight and classifying egg size from a single egg image, but the system's performance under varying lighting conditions was not extensively evaluated. Additionally, some methods require complex feature extraction processes, which can be computationally intensive and may not be suitable for real-time applications and there is a dearth of research wherein a single image of an egg is used for the extraction of egg dimensions. Building on the groundwork laid out by other studies, this research aims to develop a robust and efficient automated egg-size prediction system that integrates advanced computer vision and machine learning techniques. The proposed system seeks to classify eggs based on size by analyzing images to extract relevant features, thereby streamlining the grading process and reducing reliance on manual labour. By leveraging state-of-the-art image processing algorithms and predictive modelling, this study is an attempt to contribute to the growing body of knowledge in agricultural automation, with potential implications for improving productivity and consistency in the poultry industry. 2. Materials and Methods This study was undertaken to predict the dimensions of an egg from its image using computer vision technology. 2D images of 505 eggs were taken with a 12MP camera. The dimensions of the eggs were also measured using a vernier calliper for the length and breadth of the egg. The image dataset was manually cleaned and annotated for the egg as well as the reference notebook on which the eggs were placed, the dimensions of the notebooks were kept uniform throughout the study (21.1 cm length and 16 cm breadth). Since not just the length and breadth of the egg were to be taken into consideration but also its axes the dataset was prepared as an OBB (Oriented Bounding Box) dataset [15]. The images were structured to support machine learning workflows. Python was used for all computational tasks, with essential libraries such as OpenCV, NumPy, PyYAML, and TQDM for image processing, numerical operations, YAML file handling, and progress tracking, respectively [16]. The associated metadata for each egg, such as the actual egg dimensions (length and breadth in centimetres), was stored in a CSV file. The images used were in JPEG formats and were pre-processed to ensure consistent size and resolution for uniformity in detection tasks. The data preparation steps included image enhancements designed to enhance its quality and suitability for machine learning tasks [17]. Brightness was enhanced by converting images to HSV (Hue, Saturation, Value) colour space, adjusting the V-channel, and merging back to RGB colour space. Various values of brightness were attempted before the optimal value was obtained. Contrast enhancement was performed using Contrast Limited Adaptive Histogram Equalization (CLAHE) on the L-channel in the LAB colour space. A strong sharpening filter was applied using a kernel to enhance edges and fine details. Normalization was also attempted in one of the heuristic training procedures wherein preprocessing, images were normalized to a pixel value range of [0, 1] to standardize the input for machine learning algorithms. During heuristic modelling, various augmentation techniques such as random rotations and flips were optionally applied to increase data variability. The entire workflow was automated through a Python script, ensuring reproducibility and scalability for larger datasets. 2.1. Data Split The dataset was split into training, validation, and test sets. 70% or 80% of the data was allocated to the training set, while 30% was held out for further splitting. The remaining 30% or 20% was split equally into validation and test sets. Corresponding labels for each image stored in a text file were also split precisely. The split was ensured to be random. 2.2. Training For the training, a computing system with GPU (NVIDIA GEFORCE RTX 4070) capabilities for accelerated deep learning computations. Two models were trained one for both the classes (Training Model 1) and the other only for the egg class (Training Model 2). yolo11x-obb model was used for training both, and to select the most optimal model. The images were also pre-processed by resizing to a standard size (640x640 pixels) before being passed to the model. The models were trained using the following hyperparameters: 1. Patience: 10 epochs without improvement to initiate early stopping. 2. Epochs: A maximum of 200 iterations for model convergence. 3. Rectangular Training: Enabled using rect=True to maintain aspect ratios during training. 4. Device: Training was conducted on a GPU (CUDA), falling back to the CPU when necessary. During the training, data augmentation techniques, such as mosaic augmentation (1.0) and horizontal flipping (50%), were applied to improve the robustness of the model against diverse real-world conditions. Moreover, the model employed erasing augmentation (0.4) to further enhance its ability to handle occlusions and partial object visibility. The optimizer employed an adaptive learning rate strategy, starting at 0.01, with momentum (0.937) and weight decay (0.0005) to enhance convergence and minimize overfitting. Additionally, warmup epochs were implemented to stabilize the initial learning rate during training. Training progress, including loss, accuracy, and validation metrics, was logged and monitored throughout the epochs. Metrics such as precision, recall, mAP (mean Average Precision), and loss were recorded for evaluation and further analysis. To facilitate the measurement of egg dimensions, a reference object with known dimensions was used. The bounding box for the reference object in each image was used to calculate scaling factors, which were later applied to the detected egg’s bounding box to convert pixel dimensions to centimetres. Two methodologies were utilized to get the bounding box dimensions. One was to retrieve the reference dimensions directly from the labels (Testing Model 1) and the other was to dynamically retrieve the reference dimensions from the yolo predictions (Testing Model 2). The actual dimensions of the reference object in cm were then used to get the conversion factor. Predictions were made with a confidence threshold of 0.7 or 0.8, ensuring only high-confidence detections were considered valid. 2.3. Egg Dimension Calculation Once the eggs were detected using YOLO, the dimensions (width and height) of the bounding box for each egg were extracted. These bounding box coordinates were in normalized pixel values (relative to the image width and height). These normalized coordinates were then converted into pixel dimensions by scaling them according to the image dimensions. This was done using the following formula (Eq. (1) and Eq. (2)) used to calculate the Euclidean distance between two points in the image [18]. The scaling factors were computed by dividing the known reference object dimensions (in cm) by the corresponding dimensions of the reference object in pixels. Fixed adjustments were also applied as offsets which were heuristically determined. 2.4. Evaluation Metrics The predicted egg dimensions (length and breadth) were compared with the ground truth values provided in the CSV file. Mean Absolute Error (MAE) was measured which is the average absolute difference between the actual and predicted values. Root Mean Square Error (RMSE) to measure the average magnitude of the error, with larger errors penalized more. Pearson Correlation Coefficient measuring the linear correlation between the actual and predicted dimensions was also derived. Additionally, the Mean Absolute Percentage Error (MAPE) was computed to assess the relative accuracy of the predictions for both egg length and breadth. The correlation between actual and predicted values was visualized through scatter plots and line graphs. Scatter plots are generated to visualize the relationships between actual and predicted egg dimensions. These plots help to visually assess the accuracy of the model. Z-scores are also computed for the actual dimensions (length and breadth) to identify any extreme outliers in the data. 3. Results and Discussion 3.1. Training The Yolo training results indicated that both the training models converged similarly irrespective of the number of classes (egg for training model 1 and egg and reference for training model 2). The results demonstrated that both models performed well, achieving high precision and recall values, with mAP50 exceeding 0.97 and mAP50-95 reaching a maximum value of 0.965. This indicated strong detection capabilities for both models. However, Training Model 1 trained for fewer epochs, exhibited comparable performance while requiring less training time (10183.1s compared to 11107.3s for Training Model 2 (Table 1 ). The training losses for Model 1 were also slightly lower, and the validation losses for both models were nearly identical, although the final validation loss for Training Model 2 was marginally higher, suggesting a slight overfitting trend in the later epochs. Based on these findings, the testing results of Training Model 1 are represented in the paper. Figures (Fig. 1 and Fig. 2 ) depict the training results for the model. The combined F1 score for all classes peaked at 0.99 with the confidence threshold set at 0.861, indicating a balance between precision and recall. The models performed consistently across classes. The Precision-Confidence Curves, Recall-Confidence Curves, and Precision-Recall Curve show precision, recall remained high across a range of confidence thresholds with the area under the curve (AUC) approaching 1.0. The uniformity and balance in feature distributions indicated that the YOLO models were well-trained on the dataset ensuring generalizability and robust object detection performance. Model predictions during the training and validation are given in Fig. 3 . The shorter training time and lower training loss for Model 1 suggest that for applications where computational resources or time are constrained, a simpler class configuration may be preferred without sacrificing detection quality. In addition to using a YOLO algorithm similar to the present study, [ 19 ] incorporated a BiFPN to enhance small object detection accuracy and a CBAM to improve feature extraction, specifically targeting leaky eggs. In comparison, [ 20 ] suggests that YOLO-TLA demonstrated a 4.6% increase in [email protected] and a 4% increase in [email protected] :0.95 over the baseline YOLOv5s model, indicating improved detection capabilities for small objects. Table 1 Training Results for yolo11x OBB Metric Training Model 2 (Last Epoch) Training Model 2 (Best Epoch) Training Model 1 (Last Epoch) Model 1 (Best Epoch) Epochs 28 28 25 25 Time (s) 11107.3 11107.3 10183.1 10183.1 Train Loss (Box/Cls/Dfl) 0.305 / 0.202 / 1.576 0.305 / 0.202 / 1.576 0.295 / 0.213 / 1.608 0.295 / 0.213 / 1.608 Validation Loss (Box/Cls/Dfl) 0.294 / 0.217 / 1.511 0.210 / 0.192 / 1.670 0.210 / 0.192 / 1.670 0.210 / 0.192 / 1.670 mAP50 0.977 0.978 0.978 0.978 mAP50-95 0.941 0.965 0.965 0.965 Precision (B) 0.978 0.979 0.979 0.979 Recall (B) 1.000 1.000 1.000 1.000 3.2. Testing The Z-scores values were mostly positive indicating that most of the egg predictions were larger than average in terms of their dimensions and therefore the offsets helped improve the predictions. The actual model predictions from the two testing models is given in Fig. 4 . The actual v/s predicted values for the two models and their correlations are given in Figs. 5 and 6 . The testing results are given in Table 2 . The negative offsets of the models reveal an underestimation of actual measurements. Model 1's smaller offsets suggest it may be more accurate in predicting both length and breadth dimensions. In this study, we also used Yolo object detection algorithms, which other researchers have reported to give good results. For example, compared to detection models such as faster R-CNN Yolo was reported to exhibit superior performance [ 21 ]. Detection and Identification of Horse Freeze Branding Using a Rotational YOLOv5 Deep Learning Model was achieved with an accuracy as high as 95.6% [ 22 ]. Support Vector Machines (SVM) have also been reported to classify egg sizes into six categories, attaining an average accuracy of 87.58% [ 23 ]. Similarly, a study on hatchery egg classification using machine learning algorithms reported an accuracy of 80.4% [ 24 ]. Table 2 Testing Results for the prediction of egg dimensions Testing Model 1 Testing Model 2 Mean Absolute Error (Length) 0.28 0.29 Mean Absolute Error (Breadth) 0.19 0.26 Root Mean Square Error (Length) 0.36 0.36 Root Mean Square Error (Breadth) 0.23 0.32 Pearson Correlation (Length) 0.81 0.88 Pearson Correlation (Length) 0.85 0.81 Length Offset -0.8 -1.7 Breadth Offset -0.6 -1.5 Building upon previous advancements in automatic egg grading systems, this study utilizes a YOLO-based model to predict the length and breadth of eggs with a high degree of precision. Unlike traditional methods that focus on broad categories like weight and size, our approach offers a more detailed understanding of egg morphology, which can be applied in breeding, quality control, and research applications. By leveraging a single camera, the cost of setting up this system remains low. However, the spherical nature of eggs introduces slight errors in the testing model, particularly in the 2D aspect. Despite these challenges, studies have demonstrated that models like Segformer achieve impressive results, with a mean intersection over union (IoU) of 96.15% and mean pixel accuracy of 97.17% for small batch egg image segmentation [ 25 ]. In contrast, YOLOv11 OBB models, which are designed to handle object rotations, excel in situations where the orientation of eggs varies, offering superior performance compared to traditional segmentation models. These models are also better equipped to handle complex backgrounds, a key advantage when dealing with images that do not have plain or simple backdrops. This added robustness enhances the overall detection accuracy, ensuring more reliable results even in environments with varied visual noise [ 26 ]. This added robustness enhances the overall detection accuracy, ensuring more reliable results even in environments with varied visual noise [ 27 ]. Furthermore, research by Chen et al. (2020) has shown that utilizing advanced deep learning architectures such as YOLO in agricultural applications can improve detection efficiency, enabling faster and more accurate measurements. Recent studies also highlight the utility of convolutional neural networks (CNNs) in improving segmentation accuracy, particularly in complex agricultural imaging tasks [ 28 ]. 4. Conclusions This study highlights the significant potential of AI-driven computer vision techniques to accurately predict egg dimensions from 2D images. By automating the egg grading process in the poultry industry, we can revolutionize efficiency and productivity. The findings emphasize that integrating machine learning into egg size determination not only streamlines operations but also minimizes labor costs, making it an invaluable advancement for traditional grading practices. Declarations Conflict of Interest Statement The authors declare that there is no conflict of interest Funding No funding was received for this work Author Contribution HH conceived the experiment, AH, HH conducted formal analysis, AH, HH, PHB, and AM were involved with data preparation, AH, HH, wrote the manuscript, AH, HH, and AAK reviewed the manuscript, AH, HH, AAK, provided the resources References Roberts JR (2004) Factors Affecting Egg Internal Quality and Egg Shell Quality in Laying Hens. Jpn Poult Sci 41:161–177. https://doi.org/10.2141/jpsa.41.161 Zhu Y, Song D, Wu X, et al (2024) FEgg3D: A Non-Contact and Dynamic Measuring Device for Egg Shape Parameters and Weight Based on a Self-Designed Laser Scanner. Agriculture 14:1374. https://doi.org/10.3390/agriculture14081374 Matias Y (2023) Accelerating climate action with AI Wu R, He P, He Y, et al (2024) Egg production monitoring in commercial laying cages via the StrongSort-EGG tracking-by-detection model. Computers and Electronics in Agriculture 227:109508. https://doi.org/10.1016/j.compag.2024.109508 Alvarenga FAP, Borges I, Palkovič L, et al (2016) Using a three-axis accelerometer to identify and classify sheep behaviour at pasture. Applied Animal Behaviour Science 181:91–99. https://doi.org/10.1016/j.applanim.2016.05.026 Thangaraj JWV, Krishna NS, Devika S, et al (2024) Estimates of the burden of human rabies deaths and animal bites in India, 2022–23: a community-based cross-sectional survey and probability decision-tree modelling study. The Lancet Infectious Diseases S1473309924004900. https://doi.org/10.1016/S1473-3099(24)00490-0 Hamadani A (2024) Artificial Intelligence on Farms: Sheep Breed Classification Using Computer Vision. Indian Journal of Animal Production and Management 40:260–268. https://doi.org/10.48165/ijapm.2024.40.4.8 Hamadani A, Ganai NA, Khan NN, et al (2022) Comparison of various models for the estimation of heritability and breeding values Hamadani A, Ganai NA (2022) Development of a multi-use decision support system for scientific management and breeding of sheep. Sci Rep 12:19360. https://doi.org/10.1038/s41598-022-24091-y Hamadani A, Ganai NA (2023) Artificial intelligence algorithm comparison and ranking for weight prediction in sheep. Sci Rep 13:13242. https://doi.org/10.1038/s41598-023-40528-4 Hamadani H, Khan A (2015) Automation in livestock farming – A technological revolution. International Journal of Advanced Research 3:1335–1344 Yang Y, Wu L, Yin G, et al (2017) A Survey on Security and Privacy Issues in Internet-of-Things. IEEE Internet of Things Journal 4:1250–1258. https://doi.org/10.1109/JIOT.2017.2694844 Ab Nasir AF, Sabarudin SS, Abdul Majeed APP, Abdul Ghani AS (2018) Automated egg grading system using computer vision: Investigation on weight measure versus shape parameters. IOP Conf Ser: Mater Sci Eng 342:012003. https://doi.org/10.1088/1757-899X/342/1/012003 Thipakorn J, Waranusast R, Riyamongkol P (2017) Egg weight prediction and egg size classification using image processing and machine learning. In: 2017 14th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON). IEEE, Phuket, pp 477–480 Ultralytics (2024) Oriented Bounding Boxes Object Detection Python Software Foundation (2023) Python Language Reference, version 3.10.12 Sonka M, Hlavac V, Boyle R (1993) Image pre-processing. In: Image Processing, Analysis and Machine Vision. Springer US, Boston, MA, pp 56–111 Szabo FE (2015) E. In: The Linear Algebra Survival Guide. Elsevier, pp 101–118 Luo Y, Huang Y, Wang Q, et al (2023) An improved YOLOv5 model: Application to leaky eggs detection. LWT 187:115313. https://doi.org/10.1016/j.lwt.2023.115313 Ji C-L, Yu T, Gao P, et al (2024) Yolo-tla: An Efficient and Lightweight Small Object Detection Model based on YOLOv5. J Real-Time Image Proc 21:141. https://doi.org/10.1007/s11554-024-01519-4 Jiang C, Li X, Ying Y, Ping J (2020) A multifunctional TENG yarn integrated into agrotextile for building intelligent agriculture. Nano Energy 74:104863. https://doi.org/10.1016/j.nanoen.2020.104863 Hua Z, Jiao Y, Zhang T, et al (2024) Automatic location and recognition of horse freezing brand using rotational YOLOv5 deep learning network. Artificial Intelligence in Agriculture 14:21–30. https://doi.org/10.1016/j.aiia.2024.10.003 Çelik A, Tekin E (2024) Classification of Hatchery Eggs Using a Machine Learning Algorithm Based on Image Processing Methods: A Comparative Study. Braz J Poult Sci 26:eRBCA-2023-1882. https://doi.org/10.1590/1806-9061-2023-1882 Waranusast R, Intayod P, Makhod D (2016) Egg size classification on Android mobile devices using image processing and machine learning. In: 2016 Fifth ICT International Student Project Conference (ICT-ISPC). IEEE, Nakhon Pathom, Thailand, pp 170–173 Liu C, Wang Q, Ma M, et al (2023) Single-View Measurement Method for Egg Size Based on Small-Batch Images. Foods 12:936. https://doi.org/10.3390/foods12050936 Khanam R, Hussain M (2024) YOLOv11: An Overview of the Key Architectural Enhancements Li J, Han Y, Guo L, Hao W (2022) Research on Lightweight Pedestrian Detection Model in Complex Background. In: 2022 International Conference on Machine Learning, Control, and Robotics (MLCR). pp 91–95 Peng M, Liu Y, Qadri IA, et al (2024) Advanced image segmentation for precision agriculture using CNN-GAT fusion and fuzzy C-means clustering. Computers and Electronics in Agriculture 226:109431. https://doi.org/10.1016/j.compag.2024.109431 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6016705","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":419143950,"identity":"791ae8fd-98f5-4048-8ae5-f9a2b30f3f9f","order_by":0,"name":"Henna Hamadani","email":"","orcid":"","institution":"Sher-e-Kashmir University of Agricultural Sciences and Technology of Kashmir","correspondingAuthor":false,"prefix":"","firstName":"Henna","middleName":"","lastName":"Hamadani","suffix":""},{"id":419143951,"identity":"633d1569-88c0-4bc6-a166-2080aa982b4b","order_by":1,"name":"Ambreen Hamadani","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA40lEQVRIiWNgGAWjYLCCBCBmY+Z/+ABI8/ARpwWoh5+9h9kApIWNaGske86wSYDYBLXw8x9+9uHhD5t8gxu5xyq/5tjJsDEwP3x0A48WyRlpxjMSEtIsN9zIS7stuy0Z6DA2Y+McPFoMbvAwA/1y2MDgRoLZbcltzEAtPGzS+LTYnz+D0FIsua2esBYDhhyIFqD3zRg/bjtMWIvEjTRjhoS0NAN+9rZkacZtx3nYmAn4hb//8GPGHzY2BmzMzAc//txWbc/P3vzwMT4tKICZB0wSqxwEGH+QonoUjIJRMApGDAAAWadCifHz0y4AAAAASUVORK5CYII=","orcid":"","institution":"Sher-e-Kashmir University of Agricultural Sciences and Technology of Kashmir","correspondingAuthor":true,"prefix":"","firstName":"Ambreen","middleName":"","lastName":"Hamadani","suffix":""},{"id":419143952,"identity":"bdd04668-24ec-4c13-acfa-7aa0e6c470de","order_by":2,"name":"Pakcha Hannah Boje","email":"","orcid":"","institution":"Sher-e-Kashmir University of Agricultural Sciences and Technology of Kashmir","correspondingAuthor":false,"prefix":"","firstName":"Pakcha","middleName":"Hannah","lastName":"Boje","suffix":""},{"id":419143953,"identity":"39a4eef4-a415-4d7e-b2b2-dc858ce72164","order_by":3,"name":"Amelia Moyon","email":"","orcid":"","institution":"Sher-e-Kashmir University of Agricultural Sciences and Technology of Kashmir","correspondingAuthor":false,"prefix":"","firstName":"Amelia","middleName":"","lastName":"Moyon","suffix":""},{"id":419143954,"identity":"586c617e-0311-4069-bf1e-e5e7c836d36f","order_by":4,"name":"A. A. Khan","email":"","orcid":"","institution":"Sher-e-Kashmir University of Agricultural Sciences and Technology of Kashmir","correspondingAuthor":false,"prefix":"","firstName":"A.","middleName":"A.","lastName":"Khan","suffix":""}],"badges":[],"createdAt":"2025-02-12 16:08:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6016705/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6016705/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":77304647,"identity":"363baa06-dd80-418b-9115-59e16ee0c08b","added_by":"auto","created_at":"2025-02-27 08:55:02","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":216067,"visible":true,"origin":"","legend":"\u003cp\u003eTraining Results for Training Model 1\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6016705/v1/9759586736453bedd1112033.png"},{"id":77304635,"identity":"aeffcd35-bd34-4317-a106-7dff53981b52","added_by":"auto","created_at":"2025-02-27 08:55:02","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":232898,"visible":true,"origin":"","legend":"\u003cp\u003eTraining Results for Training Model 2\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6016705/v1/a0ba14044ad00350b978d9dc.png"},{"id":77304634,"identity":"dd2f9bfe-1318-4a07-a708-8bcfa4ba381e","added_by":"auto","created_at":"2025-02-27 08:55:02","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":274127,"visible":true,"origin":"","legend":"\u003cp\u003eTraining and Validation Bounding Boxes for the 2 models\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6016705/v1/9be49bd1293a7e6187ddae47.png"},{"id":77304655,"identity":"13704dba-a097-4b4e-9dd4-3d2583021301","added_by":"auto","created_at":"2025-02-27 08:55:02","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1570235,"visible":true,"origin":"","legend":"\u003cp\u003eYolo Predictions for Testing models 1 and 2\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6016705/v1/ddb14e98d1e50b630a6a9b7b.png"},{"id":77306928,"identity":"f5e780b1-4e3e-4244-adc0-c99bcf9159ca","added_by":"auto","created_at":"2025-02-27 09:11:02","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":161367,"visible":true,"origin":"","legend":"\u003cp\u003eThe actual v/s predicted values for the model 1\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6016705/v1/89cdd06b0057c8fe566ada10.png"},{"id":77305705,"identity":"90f57060-90d9-4bf0-8de3-b3939595c968","added_by":"auto","created_at":"2025-02-27 09:03:02","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":162493,"visible":true,"origin":"","legend":"\u003cp\u003eThe actual v/s predicted values for the model 2\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6016705/v1/42940c1ff8c491ba82fa832e.png"},{"id":77307321,"identity":"7586f34f-1557-480f-8032-e2f18daaac48","added_by":"auto","created_at":"2025-02-27 09:19:03","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3239886,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6016705/v1/c1e1295a-12c5-4df9-8ae0-b21241be83aa.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Enhancing Egg Grading Precision through AI and Computer Vision-Powered Morphometric Analysis","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eEgg size measurement is one of the most important aspects both in hatchery management and in the poultry industry. This is an important aspect that influences pricing, packaging, and consumer preference and the evaluation time between laying and consumption is a crucial factor [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Traditional methods of egg size assessment, often involve manual measurement or mechanical devices. These methods are labour-intensive, prone to human error and time-consuming. They may also cause production losses during handling [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e] and due to these reasons are also expensive in the long run [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. In addition, egg weight, as well as egg quality are important indicators of hens laying performance. They represent factors which affect the health and overall well-being of laying hens [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. The advent of computer vision and machine learning technologies offers a promising avenue for automating this process, enhancing both efficiency and accuracy.\u003c/p\u003e \u003cp\u003eMachine learning has been successfully applied in animal sciences for tasks such as livestock behaviour recognition [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], disease detection [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], and animal identification [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], making breeding decisions [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], building decision support systems [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], making futuristic predictions [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e] and for farm automation [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Recent studies have demonstrated the growing potential of computer vision systems in egg grading and defect detection. For instance, [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] developed a two-stage model combining real-time multitask detection and random forest networks to predict egg categories and weights, achieving high accuracy in sorting and weighing eggs. (Ab Nasir et al., 2018). Similarly, [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] investigated an automated egg grading system using shape parameters, finding that shape-based grading yielded better recognition results compared to weight-based methods. However, existing approaches often face challenges such as limited generalizability across different egg types, sensitivity to environmental variations, and computational inefficiencies. [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] presented image processing and machine learning techniques for predicting chicken egg weight and classifying egg size from a single egg image, but the system's performance under varying lighting conditions was not extensively evaluated. Additionally, some methods require complex feature extraction processes, which can be computationally intensive and may not be suitable for real-time applications and there is a dearth of research wherein a single image of an egg is used for the extraction of egg dimensions.\u003c/p\u003e \u003cp\u003eBuilding on the groundwork laid out by other studies, this research aims to develop a robust and efficient automated egg-size prediction system that integrates advanced computer vision and machine learning techniques. The proposed system seeks to classify eggs based on size by analyzing images to extract relevant features, thereby streamlining the grading process and reducing reliance on manual labour. By leveraging state-of-the-art image processing algorithms and predictive modelling, this study is an attempt to contribute to the growing body of knowledge in agricultural automation, with potential implications for improving productivity and consistency in the poultry industry.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cp\u003eThis study was undertaken to predict the dimensions of an egg from its image using computer vision technology. 2D images of 505 eggs were taken with a 12MP camera. The dimensions of the eggs were also measured using a vernier calliper for the length and breadth of the egg. The image dataset was manually cleaned and annotated for the egg as well as the reference notebook on which the eggs were placed, the dimensions of the notebooks were kept uniform throughout the study (21.1 cm length and 16 cm breadth). Since not just the length and breadth of the egg were to be taken into consideration but also its axes the dataset was prepared as an OBB (Oriented Bounding Box) dataset [15]. The images were structured to support machine learning workflows. Python was used for all computational tasks, with essential libraries such as OpenCV, NumPy, PyYAML, and TQDM for image processing, numerical operations, YAML file handling, and progress tracking, respectively [16]. The associated metadata for each egg, such as the actual egg dimensions (length and breadth in centimetres), was stored in a CSV file. The images used were in JPEG formats and were pre-processed to ensure consistent size and resolution for uniformity in detection tasks.\u003c/p\u003e\n\u003cp\u003eThe data preparation steps included image enhancements designed to enhance its quality and suitability for machine learning tasks [17]. Brightness was enhanced by converting images to HSV (Hue, Saturation, Value) colour space, adjusting the V-channel, and merging back to RGB colour space. Various values of brightness were attempted before the optimal value was obtained. Contrast enhancement was performed using Contrast Limited Adaptive Histogram Equalization (CLAHE) on the L-channel in the LAB colour space. A strong sharpening filter was applied using a kernel to enhance edges and fine details. Normalization was also attempted in one of the heuristic training procedures wherein preprocessing, images were normalized to a pixel value range of [0, 1] to standardize the input for machine learning algorithms. During heuristic modelling, various augmentation techniques such as random rotations and flips were optionally applied to increase data variability.\u003c/p\u003e\n\u003cp\u003eThe entire workflow was automated through a Python script, ensuring reproducibility and scalability for larger datasets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.1. Data Split\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe dataset was split into training, validation, and test sets. 70% or 80% of the data was allocated to the training set, while 30% was held out for further splitting. The remaining 30% or 20% was split equally into validation and test sets. Corresponding labels for each image stored in a text file were also split precisely. The split was ensured to be random.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.2. Training\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the training, a computing system with GPU (NVIDIA GEFORCE RTX 4070) capabilities for accelerated deep learning computations. Two models were trained one for both the classes (Training Model 1) and the other only for the egg class (Training Model 2). yolo11x-obb model was used for training both, and to select the most optimal model. The images were also pre-processed by resizing to a standard size (640x640 pixels) before being passed to the model. The models were trained using the following hyperparameters:\u003c/p\u003e\n\u003cp\u003e1. Patience: 10 epochs without improvement to initiate early stopping.\u003c/p\u003e\n\u003cp\u003e2. Epochs: A maximum of 200 iterations for model convergence.\u003c/p\u003e\n\u003cp\u003e3. Rectangular Training: Enabled using rect=True to maintain aspect ratios during training.\u003c/p\u003e\n\u003cp\u003e4. Device: Training was conducted on a GPU (CUDA), falling back to the CPU when necessary.\u003c/p\u003e\n\u003cp\u003eDuring the training, data augmentation techniques, such as mosaic augmentation (1.0) and horizontal flipping (50%), were applied to improve the robustness of the model against diverse real-world conditions. Moreover, the model employed erasing augmentation (0.4) to further enhance its ability to handle occlusions and partial object visibility. The optimizer employed an adaptive learning rate strategy, starting at 0.01, with momentum (0.937) and weight decay (0.0005) to enhance convergence and minimize overfitting. Additionally, warmup epochs were implemented to stabilize the initial learning rate during training. Training progress, including loss, accuracy, and validation metrics, was logged and monitored throughout the epochs. Metrics such as precision, recall, mAP (mean Average Precision), and loss were recorded for evaluation and further analysis.\u003c/p\u003e\n\u003cp\u003eTo facilitate the measurement of egg dimensions, a reference object with known dimensions was used. The bounding box for the reference object in each image was used to calculate scaling factors, which were later applied to the detected egg\u0026rsquo;s bounding box to convert pixel dimensions to centimetres. Two methodologies were utilized to get the bounding box dimensions. One was to retrieve the reference dimensions directly from the labels (Testing Model 1) and the other was to dynamically retrieve the reference dimensions from the yolo predictions (Testing Model 2). The actual dimensions of the reference object in cm were then used to get the conversion factor. Predictions were made with a confidence threshold of 0.7 or 0.8, ensuring only high-confidence detections were considered valid.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.3. Egg Dimension Calculation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOnce the eggs were detected using YOLO, the dimensions (width and height) of the bounding box for each egg were extracted. These bounding box coordinates were in normalized pixel values (relative to the image width and height). These normalized coordinates were then converted into pixel dimensions by scaling them according to the image dimensions. This was done using the following formula (Eq. (1) and Eq. (2)) used to calculate the Euclidean distance between two points in the image [18].\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAA3kAAAJHCAYAAAAt95+8AAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAAFiUAABYlAUlSJPAAAOsJSURBVHhe7P3tiyNVwj/+v/N7rk51fOT6lSXVwiX6oQc3PYpXz0ALdrKzsriMbjK6yIDiTKIIe+P0THqGZXHu0uo+ELdvZAZELk16p0VZTNs9Aw6YOKzaSodV/IBdQfwMPkp1Vv0Dzu/B5tR16qQqqUqn79LvFzQ6p07q5pxTp86pOnUqIoQQICIiIiIior7w/9MDiIiIiIiIaOdiJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1kYgQQuiB1P8ikYgeREQbgFUsERERbTZ28nahSqWCI0eOYHV1VV9EREREREQ7HIdr7kIffPABxsbG9GAiIiIiIuoD7OTtQleuXMHDDz+sBxMRERERUR/gcM1dxrZt3HrrrajX64hGo/piIiIiIiLa4bbVkzzbtjE4OIjh4WF9UYtarYaJiQkMDAygUqnoi13UuLvd1atXEY/HO3bwqtUq0uk0JiYm9EW72uzsLJLJZMcyBwDFYhHJZBKRSASRSATpdBq2bevRtgzz2F/QfK5UKkin0xgYGEAkEsHw8DCq1aoejYiIyFOlUsHs7Kwe3FY3v6HdZ1t18oKq1WqYn5/H3NwcGo2GvthFxp2enu4Ydzf46KOP8Nhjj+nBLpOTkxgdHcUjjzyCs2fP6ot3taNHj+LUqVM4cuQIstmsvthRqVSQzWZx6tQpCCFQLpcxNzeH06dP61G3BPO4vaD5vH//foyOjmJtbQ2WZaHRaODQoUN6NCIiohbFYhE3btzA0aNH9UWo1WrIZrNIJpP6IoyMjOD+++/fdjePaZsRO1gqlRIARLlcdoXn83nXv4UQIpFIiB1+uD1hmmZLeqlyuZwwDEOsrKzoi0hRr9dFPB4XmUxGX+QrHo+LVCqlB2865nFwYfNZpi0REVE7hULBt00wMzMjTNMUAEQikdAXO2ZmZtoup91tRz7Jk+699149CLZt48KFC3owNe8KWZaFkZERfRHQfPp07tw5TE1NYWhoSF9Mimg0iosXL2J6ejrQkIlarYbl5WU8+eST+qLAJicn1/19Q+ZxOGHzeW5uDplMRg8mIiJyVKtVHD58GOfOndMXAc3RJFeuXNGDW8gngJOTk/oiop05XNOPbdtIJpMclunj6tWrSKVSerDjzJkzME0T6XRaX0QehoaGkMlkcPLkybbDJWzbRiqVQqFQwMGDB/XFm4p5HF7QfM5msxgbG+PwVyIiamt8fBypVAqxWExf5Gi3THXkyBGMj4+jVqvpi2iX61knT51gIhKJoFKpOE8e9HAoTyUikYgz3li+x+Q1/ti2bWSzWWeCg2QyiS+++MJZXqvVcN9992F5eRkAWranWlhYwODgoLOedg23fnLt2jWMjo7qwUDzrtLS0hJeeOEFfRG18fDDD6PRaGB+fl5fBCg3Hv74xz9ueceKedy9Tvks39ubmprSFxERETkqlQqWlpZ822Nh7du3DwBw6dIlfRHtcj3r5E1NTSEejwMAZmZmMDIyguPHj2NlZcWJs7Ky4gwVPH78OHK5HOLxON566y1Uq1V88MEHmJ6eduJLsqH82WefYXl5GfV6HbFYDHNzc06cWCyG1dVVJBIJ4D8v30EI0TI0sVgs4ssvv8Qbb7yBTCaDpaWlXXNiXLlyBffff78eDAD4+9//DgC4++679UWO2dlZp5NdLBadcJk/kUgECwsLrt9sB7ZtY2JiApFIBAMDA67ZD6vVqtPh76azL5/Mvfvuu/qibdXBQ4A8rlQqGB4edt14UZfJGyeDg4OuZdvBRu97u3xmB4+IiIJ6++23gTbX4rDkEz+1TUwE9HgmklKpJACIXC7nCs9kMgJAy0QPmUymZRIQr5dMM5mMMAxD1Ot1V7h8KVVdh98EKzK8UCi4wr22t1EAhP7T06db5XK57YQQfukmlUolZ/KJeDzurEtOTCH3V8+j7SCXy4lCoSDq9boA4LzovLKyIgzDEABEPB7XfxaYPH5dIpFwTQJUr9fXVdby+bzndoJql8eWZYlEIiHq9bpzvurnijy/g05Cslk2a9+98jmfz7fk6XrKEhER9Td5LQnSXkLANqq8vuvtbNrdevYkD8273YZhtNxNOHDgANB8EiTZto3PPvus5UmbzrZtTE9PI51Ot3zbbWxszPXvILyeqHz66ad60IaQTxfD/HVKn6CuX7/ueezS0tISDMPQgx0HDx50nlQ89thjaDQaqFQqePbZZwEAlmVBCNGSR9vB2bNnnfKTSCQwNzcH27Zx6NAhpNNp1Ot1fPbZZ/rPApPHrI6HLxaLWFpawvj4uPMU6dZbb1V+tfna5XEsFsPi4iKi0Sgef/xxAMC//vUvV5wff/wRUF703i42a9/1fK7VahgfH8fS0pKTx5FIxBkyTkREpJPXiI1oL/300096EO1iPe3kAUAmk4FlWa5hex999BHi8bhriN/Vq1fx9NNPO//2c/XqVQDAz3/+c32RZ1g3dsNELZcvX3Y6237kuO5OHnjgAQDA73//e1iWhcXFxcAvCHeiv9sZ5C/MrFIPPvgg0NzO2NgYpqamelbRfv/9987/p9Pplg67EAKLi4uu33jxepc1EolgfHwcUN43Vf/0IYp+guSxvLHw+eefu8I/+ugjJBKJdc/K6Xd87f6CHt9G7zuUfI7FYi35K/+IiIg22/Xr1/Ug2sV63sn77W9/CwB4//33gebdbsMw8Mc//hGNRsPp6L3xxht46KGHXL/18t133+lBFJJt21heXg7UwA9CNqQbjYbzBKVXFhcXWxrMnf6OHz+ur8aX7KCaprlt36E6fvx4yzEKIZDP5wGfJ8JBOo9hxONxLC0tOf+2bRvFYhGnTp1yxeuG3/G1+wtzfBu570REREQ7Qc87eUNDQ85TO9u28fLLL+Opp55yOnRvvPEGarUa9uzZ07OnPzuF/nQiyJ/X7KA62YgdHBz0nEL3k08+QTwe71l6qxOXeHXwbNvG5OSkMxHGwMAAJiYm9Ghb4quvvgIADAwM6Itc2qXnbjA8PAwoQxMvXbqEsbEx1/DhYrHoevKaTqe7mrym14LsO5oTsiSTSdcIAyIiop1K3sgmwkZ08gDg6aefdqYbr9VqiMViiEajSKVSWFpawssvv4xHHnlE/5mne+65B2gON9zp9KcTQf70hqluYWEB9913H9544w1YloWXX35Zj4KPP/440PuL6tMPP9VqFaOjo0ilUrAsy7MTdOnSJVy+fBlzc3MQQuDEiRM4d+7cljemi8UiXnrpJcTjcd+PjFYqFaTTaZw6dQqWZYXqGN9222160LYTJI/RvFmD5tBE+e6Z+tFW+bmTU6dOQQiBcrmMubk5nD59WlnL1ui077VaDdlsFr///e+xtLSE22+/Xfl1Zzshn4mIaHuSM9F7tZ/W66abbtKDaBfbkE7eo48+CgA4efIknn/+eSf8ySefBJqN7XaTgKjuu+8+GIaB5eXllk7CDz/84Pq3l7An0eTkJAYHB1EsFp0nFHK7cqr9SCSCWq3m/Lvbadl74b/+67/wySef4K233oJhGJienm55mjI3N4f//u//doXp5Kcn2qlUKhgdHcXU1JSTl16T1hw/fhyfffaZ00F66qmngID5tVFmZ2eRzWbxzjvvYGxszLeDeuPGDfztb3/DG2+8oS/yJdM7TIdwKwTJY0lO7Xz9+nVks1nk83nX8Y2MjGBtbc25CTEyMoJ4PI61tTUnzlbptO/ff/89jh49GnqynZ2Sz0REtH3Jm+7qe/zr9emnn8IwjJ68d059RJ9us1dSqZQwTVMPFoZh+E5jvrKy4kzFb1mWE14oFJzwfD4vyuWyyOfzzjS0qVRKlEolIZTPNeTzeVEoFJyp8+XnFmQ8fXsrKyuu7ch9lNPWy/1JJBLOceXzed9j2Qq5XM45dsmyLIEAU/XK33p9ssErXeTnCNSp69U8UxUKBWEYhu/yjSTLA5Rp9eV0+vl8XliW5XxSQVUulwUCnB5qOmy09X5CoV0eewEgDMMI9EkAWc7U82srBd33oOmxmflMRET9S7Y9Z2Zm9EUusq1immbb9pO8PumfLyPqvsXYQalU8izAmUzG8zsesgGr/qmNr1Kp5HTUTNMUpVLJ+UaV2rC0LEuYpikMw3A6hPp65ffL9HDZedQ7p9C+uyX3Yzt18ITS0Fa/KVgoFAI1TGU6eVUSsnOgL5NpKPPDy8rKijBN0zPPN4P8pp/+3TT5TZlEIuG5b0E7ebIS9irrvbbeTl67PPYib6J4pY9KfitRT+OtFHTf5XnfyWbmMxER9bdEIuF5g1mC1j6FdgNfJW/Et+sI0u7UfYuxT3l18kzTdDVgZWdqu3XyhPLkSu5vKpXyrRh0pmm2HPt6bHUHbz2CdvJSqZSrU73dBc1jy7KcGyXtbMcOXtB9FyE6eTstn4mIaPuST/N60TGTDy6IdBvyTl4/UidnME0TlmVheno61DfaNsOf/vQnAHCmi79y5Urgd7Hk5C3qR+u7Va1WcejQIbzzzjt9O0a8Uqlgbm4OJ06c8JxldDsKkse2bSOVSmHfvn1tP09h2zaSyST++Mc/Bn7HdqMF3fcwdmI+ExHR9jU0NIRCoYBcLqcvCkXOwN6r6x31Gb3Xt9vJIXHy7kqhUHA9+VD/Xz7t2U5PMYT2XqJhGPritnK5nDAMY11P3+r1ujAMo2W47U660yTLgd8wVPmeZ7vhFtuVVx7n83mRz+fFysqKiMfjIh6Pd3xqpd89rNfrgYYG91o3+y4CDl/dyflMRETbW6FQ6HpU2MrKikilUoGud7Q7sZOnkcM15Xt3aj9YvsclT0g1znbqwKjvIXbTOJWdw27fP1InO1H/tlMa+ZGdO/1P7bCWy2VnAqGdWrnqeSzLNprvdHbq5KuT8ah/W9HJC7vv6vmh/unlsx/ymYiItrdyuRz6YcHKykrXbTTaPSLiP++lUNPk5CRef/11rK6u6ot2lGQyiaWlJczMzODo0aP64o4qlQpee+01mKaJs2fP6ot3rdnZWbz77rs4depUx28YbndqHl+5cgXLy8vIZDJ48cUXd9SwxOHh4Z7vez/lMxEREe0+7ORp+qWTV6lUsH///tAf9CYiIiIiop2NnTzF5OQkxsfHgebkKju9o0dERERERLsPO3lERERERER9hJ9QICIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1EXbyiIiIiIiI+gg7eURERERERH2EnTwiIiIiIqI+wk4eERERERFRH2Enj4iIiIiIqI+wk0dERERERNRH2MkjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfiQghhB5Iu1MkEtGDiKjHWOUSERHRRmMnjwAAlUoFR44cwerqqr6IiIiIiIh2EA7XJADABx98gLGxMT2YiIiIiIh2GHbyCABw5coVPPzww3owERERERHtMByuSbBtG7feeivq9Tqi0ai+mIiIiIiIdpCePskrFotIJpOYnJzUF3WtUqlgYGAAExMT+qJAKpUKstksksmkvoiarl69ing83rGDV61WkU6nu86LfjU7O4tkMolKpaIvcrFtG5OTkxgeHkYkEllXud4ozGNvQfO4UqkgnU5jYGAAkUgEw8PDqFarejQiIiKged2YnZ3Vg9vq5je0+/Ssk1csFvHee+9haWlJX7RlKpUKPvjgA0xPT+uLSPHRRx/hscce04NdJicnMTo6ikceeQRnz57VF+9qR48exalTp3DkyBFks1l9sePSpUu4fPky5ubmIITAiRMncO7cORSLRT3qlmAe+wuax/v378fo6CjW1tZgWRYajQYOHTqkRyMiIkKxWMSNGzdw9OhRJ8y2bWSzWedm4eDgYEuHbmRkBPfffz/S6TRs23YtI3KIHiqXywKAyOfz+qItBUAkEgk92Nd22/+NZpqmKJfLerAjl8sJwzDEysqKvogU9XpdxONxkclk9EWe6vW6ACBmZmb0RZuOeRxM2DyW6UpERKQqFAoilUq5wuQ1xjAMkUgkhGmaAoAA4HndmZmZCdW+pd2lZ0/y+kWlUsGHH36oB/etWq0Gy7IwMjKiLwKa6XHu3DlMTU1haGhIX0yKaDSKixcvYnp6uuWum5erV6/CMAw89NBD+qLAksnkuociM4+DC5vHc3NzyGQyejAREe1i1WoVhw8fxrlz51zh8/PzGB4exjfffIPFxUWsrq5iZmYGADA9Pd0y/F8+Aezla1LUP9jJU1SrVfz617/Wg/va1atXkUql9GDHmTNnYJom0um0vog8DA0NIZPJ4OTJk22HUFSrVZw6dQrXrl1DLBbTF28q5nE4QfM4m81ibGyMQ1+JiMhlfHwcqVSq5fr/ww8/YGpqyjVHwtGjR52bhV6vRB05cgTj4+Oo1Wr6ItrlNqyTJ8cUywkmvN47kpM8RCIRRCIRpNPplkLabjKXSqXiTGLh9edlcnLSGeesTi6xsLCA0dFRNBoNLC0ttV1HP7l27RpGR0f1YKCZP0tLS3jhhRf0RdTGww8/jEajgfn5eX0R0EzXQ4cO4Z133tnyJ2fM4+50ymP53t7U1JS+iIiIdrFKpYKlpSXPttfx48f1IADAz3/+cwDALbfcoi/Cvn37gOZ7/0SqDenk/fDDDzh9+jQefvhhlEoloNnoUe96y4buk08+CSEESqUS5ubmMDY25sRbWFjwncylUqlg//79eOyxxyCEQLlchmEYAIBCoQCvL0NMTEzgjjvuwD/+8Q/E43GcO3fOmS3v4MGDWFtbAwAkEgkIITzX0W+uXLmC+++/Xw8GAPz9738HANx99936IqDZkZ+YmHA68uowgmq1isHBQUQikbZPO7ZKpVJx9k+fSEMeU7czTB48eBAA8O677+qLtlUHDwHyeHZ21kkn/UbL5OSkczNET8OtJsvmwMAABgYGsLCw4FqeTCadffe6AdVJuzxmB4+IiPy8/fbbQJvrbjter3fIp4Fzc3P6Itrt9Jf01kNOvBKPx0W9XnfC8/m8AOCa3CORSLRMcJLL5VomovCbzEW+mKqS29FfTgXQMqmE33rDTtKyXvKF2jB/7SZJCaNcLrekoSqRSAi0KSK5XE4UCgVnAhH5AvHKyoowDEOgWRa2G8uyRCKREPV63SlzsmxkMhknnUulkv7TwOLxeEva1et1YRiGK/9KpVJLGQwjkUisq7y2y+NSqeScS/Llb8uyXHFk+hUKBVf4VpNl07IsAUCYpqlHcfJIP6agvPI4n8+35Md2PAeIiGhryGuH2k7uJJPJtFxbVPJazsnTSLUhT/Iee+wxz2+uffXVV4AyRCyRSLiW33HHHYDP3XHd8vKy84haeuCBB4DmZCK6ffv2eT45+eKLL/SgTSWfGIb585skJazr16+3fQ9raWnJeTrq5ezZs0in04hGo0gkEpibm4Nt2zh06BDS6TTq9To+++wz/WdbLhaLYXFxEdFoFL/85S+B5rHOzs6iWCyiXC5DCOE8remGLP9qWTx9+jQajQb279/vPEX61a9+pfxq87XL44MHDzpPo5555hkAwP/9v//XFafRaMAwjLblaCvIshmLxZBIJGBZVssT5Uaj4flORFB6HtdqNYyPj7uGe0ciESwvL2u/JCKi3UpeE7zayV5qtRqKxWKg0SE//fSTHkS72IZ08vz88MMPAICvv/4aALB3715XY+jYsWMAgNXVVdfv/PjF27Nnjx7k69///rcetGtcvnwZBw4c0INd9I60nwcffBBoDoMbGxtreXG4W5VKxVVGgv4FJTvMly9fxsmTJ3Ht2rWedaIB4Pvvv3f+f2pqqqXDLoTwHYOvUocXqn9LS0stnQr5pw+v9BMkj+UNlC+//NIVfuXKFZw4ccIV1g1934P8dfowuSTLpqx3oMwq+9xzzykxuyPzOBaLteSt/CMiIupGLpfD1NRUoBuS169f14NoF9vUTp703XffAQDq9XpLY0gI4dt5U6VSKViW5XpS8uOPPwIAHnnkESUmebFt2/NpaLdkJ8A0zUB3m4IaGRlpKR9B/sJIJBJYXl7GtWvXPJ/2bgeLi4stxyiEQCKRcL1Dqv4F6TwGdddddwGA6/MilUoFa2treOqpp5SY3dH3Pchf0M74PffcA2gXv/n5eSQSicDrICIi2mzFYhEDAwPbbrQM7Qxb0smTswN98skn+qLATp48CcMwnAldbNvGn//8Z8Tj8R11MuhPJ4L8BXmCYds2isUiBgcHPYevfvLJJ4jH44HuDAUhh+IODAzoixy2bWNychLDw8P6oi1j27ZzU+Gmm27SF7vYto3BwcHAT8j6STQahWmarhswZ86cwfnz51ue2LYrd1vhv/7rvwAA3377LdDMxwsXLiCfz2sx/zPRTCTEk2AiIqKNUK1W8a9//SvUjXN5w50IW9XJk7M5vvnmm/oioNlI7GRoaAjnz5/Hnj17cOutt+LOO+/E8PAwFhcX9ajbmv50Ishfp6cPCwsLuO+++/DGG2/Asiy8/PLLehR8/PHHGBsb04NbeM1sqisWi3jppZcQj8dx5coVfTHQjPPEE0/gwoULLZ2CrWLbNpLJJOLxOADg008/1aO4nD59GpZl6cFt3XbbbXrQthMkjwEgHo87x18sFrG6uup8iBXNJ3vpdBqnTp2CZVk9u4GwXrFYDIZhOJ3O06dPI51Otzy1rVarOHnypCssiJ2Qx0REtD3INke7G6HVahV///vfQ39ntdPNatpdtqSTNzQ0BNM0MTc31/JphWKx6AznbKdYLKJaraJYLEIIgbW1tZ69B6ZqdxJuV//1X/+FTz75BG+99RYMw8D09HTLpBNzc3P47//+b1eYTp8Yx8vs7Cyy2SzeeecdjI2NtQyhlX744QcsLi725P2tXqjVakgmkxgeHsbf/vY3AMBHH32kR3MsLCygVqsFShM0O5BQpjberoIeDwDce++9QDMtstks3njjDdfyGzdu4G9/+1tL+Hawb98+LC0toVKpoFgs4sUXX9Sj4Omnnw5VPndKHhMR0fYhb7Cr7+yrqtUqxsfHPTt4s7Ozrs9VSZ9++ikMw2i5eUm7W087eXLI3ocffujqVMgZLNWZLN955x2nA3LnnXcimUxicHAQr7zyiusdH7nOy5cvu9Z56tQpFItFTE5OOn/FYhGVSsXVyZAnw6effuoKl+tdXV11rdc0TSwtLWFhYaHr76RttVgshmg0img0ikwmA2gfyZSTTtx3333Kr1r94he/AJpPaHTyQ/fHjh3D1NQUhoaGnE7j/Pw8arWaa9is+sRnK9VqNUQiEZimCQB48cUXEY1GEY/HUSwWYdu2M8umZNs2nn/++cBDJuzm+45hOlBbpV0e6+QwkN/97nfIZDItT5TlTKvbkZx85de//jX+53/+p2U/JycnMTY2Fnioy07KYyIi2j5++9vfAko7VFWtVjE6OuqMNFL/BgcHcfHixZaOnG3baDQaTnuPyKF/U6Fb8hsd6p/8Fp36p37nY2VlxfmdYRgik8m4vhvitU5Jfp/L769QKDjfzVP/5Hes9HCpVCoJwzCEaZrr+k7adiG/E2YYhpO2hUKh7fdWJJl/uVxOX+R8p1D/PppM20Qi4fm9Fq/viG2mUqnk7J9a1mS+G4bR8t26RCLhlIWEx/cddXIb6vceN0pind/Ja5fHOvk9RNM0237fR65zO5F5Ir/lqCqXy8637ILu+2bmMRER9ZdEItFyPbIsy/nGsN+f1zWnUCgIrOObr9S/IkKEnIpwm6hUKrh+/ToeeOAB3LhxwzXE84svvsCVK1ewtrbm+s1ulc1mMT09jUKhgHQ6jXQ6jXvvvTfQ7IuDg4NAm89VhDU5OYkPP/xwx7w7OTs7i2vXrjlP9pLJJB588EHcc889+PLLLz3TMJ1O48qVK/jmm29anhhtR0HzeGFhAb/61a9QLpdbnuKpKpUK9u/fj+1UtUxMTGB6erolT2zbxn333Yd33nkHQ0NDrn0fHh72/c7jTstjIiLaPqrVKvbu3duT99dlu8SrPUK7nN7r2wnkE8B2Oi0PI5FIiEwmIzKZjHM3Rd4xUe+wqP/u5fbXSz7NM01TCCGEYRieT9m8yCcbXnePurHVT/LCandXzeuJnkwvr2XbVZA8XllZ8XzK6SXo07DNIu9ylstlfZGYmZlpyVf1z8tOzGMiItpeCoVCy9O8sMrl8o5qU9Hm8m7FbHOpVMoZXuVlZWWlZw0wtWMnhyYmEgmnwySEEKZpOidZJpPp2bZ7SR5HPp8XhmHoi9vK5XKhOoZ+6vW6SKVSwjCMHTusIJFI+A5trNfrwjTNdVfaW8ErjxOJhCiXy85Q1kwm4/qNHzlMeiuHO8sOmtwXfVixH9mB8xuOupPzmIiItpdCoRD42qpbWVkRqVTK93pFtCM7efJ9PNM0RT6fF6VSSZTLZedk6faE8SOf5EmyIah2VGRHcDt28ISyz/B5L6kT2Tls97SnnYTHe5CyIb5TlMtlYZqmZz6Xy2XP90p3Ej2P1XyKx+Mdj0t2qPQ/rydoG0kt6wAC1wf1et31rq+uH/KYiIi2F9l+DWNlZaXr9hjtHq0tmR1CTh6iNuZSqdSGPD3QO3ly+KPaydsJQ7hkenVbMZTLZZFKpXyfZO1WMzMzzlOvnU7m8e9//3uB5oQ927lMe5GTovRy8qR+ymMiIiLqfzt24pXNlEwmEYvFnCn0a7UaTNN0Jpao1WoYGxvDG2+8gf379zsTnGw3clKJXrzoS0RERERE2xM7eQEkk0msrq46sw9ms1kAcDp9g4ODzrJisYjDhw93nIGQiIiIiIhoI/T0Y+j9bHBwEJFIBJFIBLVazdXBsywLk5OTAIDDhw8DAPbv3+/6oDYREREREdFm4JO8APThmkRERERERNsVn+QRERERERH1EXbyOkgmk1haWsL09LTzLh4REREREdF2xeGaREREREREfYRP8oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1EXbyiIiIiIiI+gg7eURERERERH2EnTwiIiIiIqI+wk4eERERERFRH2Enj4iIiIiIqI+wk0dERERERNRH2MkjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+INlWtVsPExAQGBgb0RQSgUqkgm80iEom4wmdnZxGJRLCwsOAKJ2BychKDg4OIRCIYGBjA5OSkHmXD2baNwcFBDA8P64u2HMtO7xWLRSSTyS0pa1tF1k3JZFJftCtVKhUMDAxgYmJCX7SrbEbd59du8Aun/2Anj4g2Ta1Ww/z8PKanp9FoNPTFu16lUsH169cxPT2tLyIf2WwWX3zxBVZXV2FZFsbGxjA+Po5qtapHpW2sWq0inU4jEokgEolgeHgYlUrFFUcu8/sbHBxEOp3G7OwsbNt2/baXFhYW8N5772FpaUlftCNUKpWWtGv3l0wmUalU8MEHH7Buok1n27Znu8EvnBSCiCigfD6vB3Xk9ZtEIiFY/fgzDGNL06dcLotyuawHbzsrKysCgCiVSvqiHWunpH0vraysCMMwhGmaIpFIOOUfgCgUCi1x5bKZmRknvF6vi1Kp5NQthmFsaLkol8sCgGf9tt2Vy2VhGIarnMnjSSQSTli9XheZTMYVpseh9duIMlSv113nRz/wazf4hZMQfJJHRIFUKhV8+OGHenBb3fyGgH379ulBm+rMmTN60Lb0z3/+EwBw880364t2rJ2S9r10/vx5nD9/Hqurq1hcXMQ333yDTCYDNJ/UqoaGhpBIJAAAd999txMejUZx8OBBLC4uIpPJoNFo4Fe/+hWf6PqYmprCyMiIHuwSjUYxNTWFWCymL6IesW0bFy5c0IPX7dKlS/jhhx/0YNpl2Mkjoo6q1Sp+/etf68FtdfMb2nrZbHbHDEPrt0bMTkr7Xrr33ntx9OhR59+yc2GaJhqNRsuwzU6mpqaQSqUAAIcOHdIXU8gbSUNDQ3oQ9YBt20gmkz0falgsFjE+Pq4H026kP9rrlnxcqv7p4RvxiH8rtiuHNah/6uP2fD6/IdvebtvVw3s9xEg9Hq9tqHnc7XAHOaTBNE1RLpdFoVAQhmGIeDwu6vW6EyeTyThDiOLxuGs/9HWUSiVhmqaTDysrK05cy7JELpcThmEIIYTI5XICgMhkMk6clZUVkUqlnGNLpVLCsixnuRBCzMzMuPYnlUq1pH+n9ej7ou+3PP5SqeQaPiX/2un0G3V4hd92Vfl83oljmmboYSj1er1lHV5lxivN9HSVgsQtl8vOcKd6ve4ctzoETV9PJpMR8XjclV56GZPU9YtmOsl0z+VyTjw1vlw3PMqyV32qppO+r3qZEs1tqOmcyWQ807odr7xSy4VX3YCA9Z6eJ5lMRqA5vM9vPzvtj/DIC+GRbysrK04am6bpqhvapb1lWc5ywzBEKpVy1Rnt6Psg6fvbqey006lcdUses5pOani7bajDOoMM29TrQzU9MpmMZ36reSRZluU6R7zqq0KhIBKJhFOO1DKoD08VAa5BvSCPp9M5pMYJUmb0c0dPi6Blu9N62lHTWz8HvcqGV3p75Yu6XkkvR37XN8uynHD1T+Zrt/WoV93oVbeo+SzzXv6JkGmm1yUqy7Ja0tLr90HqD7kPOj087LH2s54e4czMjJNwagbBoyLspa3Ybrk5ph3aewFSLpcTqVRKD163rdquZVnOSaieMPl8XsTj8ZaLcK+oeeu1jXg87pkOQc3MzDjHlc/nRSqVcirWlZUVUa/XRTwed8rRysqKa7llWS0XOlkxyvUahiHq9brT0ZBx8/m8U/nJC4Jcv6wES6WSQLNilRcHedGQjetCodBS9jutR9+XQqEg8s2Ou2xw6OeOnvdB+P1GVspBtpvJZEQqlXL2W/42aL7LPFQbarIRpjZMZBrJ9aplXr/AB4m7srLidOITiYTTeEGzcyTjGIbh7Ju6HigXILWcynzW15/L5UShUHBdMPUyAeWGgty2LJ+SbCSov5Xx25Up0UwHQ3nXp9xsqOh52k48HhfxeNwp37IOiCs3XiS/ffVTKBScvE8kEk66yXSER/kLsj96Xkhqvs3MzIhMJiPK5bKzDv3c8DsetQ6SjWL9t356UXbaCVquupFIJIRpmnqwUwd02kdZV3t1GlTlctlVBnK5nNOolnWknt5lj05evV4Xpmk6ZUPWPeq+lkolV/2TyWREqVQSpeaNMT3d5DrkdvRrUK/I49GPUxemzASpu4OU7SDr8ZNXrsWZ5g20fPNaL/Nb3W9Zh8k6ua50wtVypOaj3P96F9dVeSyq9dajXmVTvbZ4tRFlu0IeQ9A00+sSlawLZD7J8wPKjZcw9YdXWnmFhznWfteaWuskM1smrLywbbSt2K68IHtVNPEN7Phs1XZlxQOtA7RR25PkCaxvp16vC6PZOVoPmZ6ygqrX68628vl8S8UlG7dq5eF1ERDN/IBW2crjUS9ssgGZ0O4KCqVsy/xOpVIip901zeVyrgtVkPXIePC4062mR7uwTvx+I7erd570+OXmTQ290YNm5zkIedFV1yEvgjK/ZFnS88+yLGdb8vdh4grlrqJ6UZTLzeZEEyq5HmgXM1lO9YaU3KYsQ8LnIu/1e1km1DCveCJgmZIdClWpVGr5nR+5bfVYhHJ+6evx29d21HxS6xR5Xqv5F3Z/4FHe5Tr0c1ZvmAif45GNIP0c0LfTjtd6RYiy047Xur3KVTf8GrYy7TqtX8YLmlayAarWS+p1T92eVzp5hcly5RUvoY1c8ErLfMBr0Hqp+9SOLDPqueN13OUAdXeQsh1kPZ3IG6Gy8y15dU4SPjcWZNnoVAZEyOuqVz2w3nrUb79kuFe5yeVyrrZBmDQTbY5ND5NlXJ5jXmXer/7wSiu/8DDH2s96/k7e2bNnEY/HMTc3h4mJCVy8eBFTU1N6NNRqNWSzWaTTaX1RV4JsV45/jjS/pVQsFl3Lw3rqqadgGAZeeuklV7h80Vsdx16pVJBOp3vyTZ2g261Wq863owYHB9f9AnosFsP//M//AACefvppPPHEE/jLX/7iO15ffjtlvcd86tQpoPmtJ9X8/Lzzcr6q2+3K7USjUeeYLly4gN/85jeueD/72c8AAHNzc07Yz3/+cwDA448/7oQBwF/+8hcAwOXLl13hAHD8+HGgma6xWAzVahVLS0vOpALSHXfcAQB49913AQADAwOYnp525edvf/tb5/+Drkd18OBBPWhT3g3yOv8//fRT5/9fe+01jI2NIRqNOmHRaBSGYQR+V+fChQtIp9OudYyMjEAI4dQRV69eRaPRwMMPP6z88j95IydxmJ+fDx1XSiQSziQHQ0NDiEajWFhYgGVZOHLkiCtuLBaDaZqusE727dvnOTmCOunNHXfcAcMwcNtttzlht9xyi/P/7QQtU7fccguWl5dddevBgwcDb+f1119HPB5vOZY//elPQDMv10uue9++fa666+DBg4jH42g0Gvjkk0+AHu/PL3/5Sz0IaF4b2rnpppsAAKdPn3bCotFoS73UrSBlp531lKt2isUiDMNw6snNMDg4CGj1kjyvAeCDDz5wwr3cdNNNMAwD99xzjxPWblKgBx980FUvSdevX3f+P+g1aDPp5470xRdfOP8fpO4OUraDrKeT22+/HQDw2GOPudbzhz/8AWhe62zbRq1Ww9LSkvM+p0pOiPTaa6/pi3x1e11dbz3qZ2RkBIlEAnNzc6jVaq5l09PTePTRR51/B00zP/KaoV/fjh8/DiGEc45tVP0R5lj7Wc87eQBw8eJFAMC5c+ec/9flcjkUi0X8+9//1hd1rdN2//rXv+LUqVMQQuD8+fM4fPjwujo+0WgUJ06cgGVZrpNxdnYWTz/9tCvukSNHelYhB93u+fPnceXKFdTrdYyNjWF0dNRZ1q2DBw8ilUpheXkZe/bs8azEpNOnT8OyLD04NHmyTk9Pu07Wl156ydW5kXq13VqthkajgWPHjrm+GbR3715XnHZk+iwvL+uLWnz99dcAgL1797q2d+zYMQDA6uoqADgTFOzduxfZbBa1Wg1DQ0NOJyLoerYr9SX05eVlzM3NtXy3Sca5ceOG8stWlUoFjUbD6YT7ee+99wCfBplszMiOTJi47bz//vuAcjFVycZmL6XTaaytrSEWizk32YJ2UoKWqYceegjxeByHDx92vq0Fpcy2U61WYVmWZ6M3FouFatR167HHHgMAfPnll9tif6B0MqanpzE4OOjU+UHSdDOsp1z5sW0br7zyiud1fCvIDvrnn3+uL3IZGhrC2toaDh48CNu2MTk52dLIDaMX16DNpLbngtTdQcp2kPV0KxqNOjeuvv76a1y9ehXw6WTcddddwCZ1rNdTj3Yib2a//PLLTlixWGzpSPvR08yP7Mx6Xd9UG1F/SOs91n6wIZ28oaEh507I+fPn9cVAM6FPnDihB69Lp+2ePXvWaQjLk+Wnn37SYoUj7wbIwmTbNorFYstdgtXV1Za74OsRZLvFYhGxWAzRaBSPP/64q/G8HidPngSalZ1fJ3lychIHDhzQg7smL5TyZK1UKjAMo+VuYi+3+/333wMASqUSmkObW/687oB367vvvgMA1Ov1lu0IIZyG9NDQEJaXl52Or2marqeWQdezE1iWhVwu13IM8s/rSWA32t1sUqdpR8i47WxF46xWqyGZTGJsbAwHDhzwfBLuJWiZikajWFxcRC6Xw9LSEvbv3490Ot32jq/UqS4OMxtgL2yn/ZmamsLMzAzW1tZw+PBhDA8P+9a9W6HbcuXn9OnT+OMf/9hSv4cly+UvfvELfdGGsW0bExMTuPPOO4F1fhJjs69BvRS07u5UtoOupxfazda73rIYxnrq0U5GRkZgmqbrpvkbb7yBJ598Uo+6aXpdf0jb8Vg324Z08hYWFjAwMOAMn1xYWNCjbIgw263VajAMw7k70y15J0o+Vbt69WrL0LCNEHa7N27c6Fknc3x8HPl8HmgO29QtLCzghx9+6Gnlm06nXSfr22+/jT/+8Y+uOL3erhxK8vHHH+uLQgsy/E7ePZRDxdqJxWJYXFxEuVxGPB7H+Pg4JiYmgJDr2e4Mw8CVK1f04NCCDj376quv9CCH3pgKE3c7qFarztDD1dVVpNNpzzvWXsKUqWg0irNnz8KyLKRSKczNzSGZTOrRfKnDdb2ow3o2ihyGim2yP2jemPzmm2+Qz+exvLyM0dHRLblRoFtPufJSLBZx4MCBddfj8kks2gyVDavTeS1fCbly5Qq++eYbHD9+vOOTjHZ6eQ3abGHq7nZlO8x61kM9j9Vhp7og1/JeWG892o688fDyyy+jVqthdXW17agsP0HqPnXosZde1x+6Xh3rTtXzTl6tVsOrr76KF198ERcvXoRhGPjd737XkzsQ7YTdbjabxT/+8Q/fTlEY8v2MV155Ba+88krLuzobJeh2bdvGqVOn8NZbb+mLQpuYmMCRI0dw/PhxZDIZLC8vu54iVatVvPnmmzh79qzrd72gnqzT09N46KGHnGUbsd2hoSEYhoG5uTnPcqQOlfUjfzc2NqYvanH//fcDAN588019EeCzvZGRESwuLsI0TZw7dw7ocj3b1djYGJaXlz2fWtRqtY5D5eST+6WlJc8GsXzPU74Dcu3aNS0G8OOPPwKA84Q4TNx25NOFTu/59Ir8Xpj+rnIQ3ZSpWCyGYrHoDO8OkldyCKRXftu2DdM0Oza010Peyd+3b9+22B9dNBrF8ePHMTMz4/nu51ZYT7nSyXK03g4emjcjob0P262g5/Xp06exvLyMixcv9qRt0Ytr0FYJW3f7le2w6wlLPY/ljXCvTmWYa3kvha1Hg1Bvmr/88sue7yC2E6Tue+CBB4Dme81e5LW3l/WHl/Ue607X006ebdvIZrPI5/OINievOH/+PBqNBp599lk9es+E3W42m8WRI0fWXfFLsVjMOQEbjcam3SUIsl3btvHEE0/gnXfeWfdFp1gsotFoOBfgF198EaZpYnx83KmAl5aWXOPn0bzYhp0ExUs6nYZhGJienkYmk3EdTy+2Ky/kqnQ6DcuykEwmXZ2EarXqvJul0tchx/jrE7JAuWhIQ0NDME0Tc3NzyGazruXFYtEZMqcvi0ajeOaZZ5x/B13Penl1mjoJ+5tHHnkEADA6Ouq6uMlzPsiTeDn0I5VKtaSFLLePPvqo05jS9/H999+HaZpOuQ8TV9LzGsrTBX0SHfjEXy+v91TbDU+Sas13PoOUqcnJyZZjee6551z/bkcO4dcnWarValheXl7X0DedVxrPzc0hkUg4jZfN3B8vshGrb78X7+b0SrflSlcsFvGvf/2r5dxBs84LQ35U3jAMZ9RJGHrZeP/992EYhuvGohe9PoDHNSGsoNegXnR4eilI3R2kbAdZT1B6uZTnsbx+Dg0NIZFIoNFotHSg5bVc3lzfKLVabd31qM627ZYy/cILLwDN689TTz3lWqbqlGZ+5FBJy7Jazt9sNus8retV/dFO0GPtS/p0m92qN78lok+XKqe6hcf08rlcznOaWMMwhKl8B6ydsNvN5XLOtLbyt6KL7er8pqxVxePxlmm05fSx+v4H1W678vjkVMcrKytOvLDbldMY61Payqlu9SmVJX3f5FS3XvsbhNd0u16CbqNerzvffpHf4dGXy88gyPSKx+Mtxyv3K658xFOmmZrnlvIBVK/9k9+IkWkqp3NWpzGW+yC3L9epbifIeuoe36yRv5XHqx6jGjeXywU6T7x+E3a7crp6mb6yDOmfX/BTb37yoF1aCG3KaHlsM82PzuvlLWhceX7Cp8zKYzOaH0EuN7+nJNMnl8s5aSHLqVqfyTQztGnwZ5rfYTOVbwHJdMs0v9WWUT66nsvlnPIojy2VSrmm7Q5SpvLNz1WodWwqlWqp5/2o55uc4tpqfvNInwpbjasv60TmSb75MWpZV+rpGGZ/1PKrrkPmm3p+queA17TlatrLMjQzM9OSzkHOQdGDstNO0HLVjjzuRHPKdfXP0OpQNZ3Vc6per4tSqeTsj1pHBiV/m0qlXOc1PKbDl+F6+Zdh5XJZ5JufP5DHNjMzIyzLcn6b0D6hoF6LJLUMyt/EPa5BsjwFrRd1ctv6elVhy0ynujto2e60nk7kdmQ9K5TzWK+bZD2n1uXl5rVcn3bfqwyo53aY61s+nxeFQkEUmt/WW089ajU/E2OapnNO6urN62LCpw0YNs3ksan5Jtch90Wez+r+BK0/1HRV890vXNXpWPtZzzp5MqNkYZVkmPyTiSwrQ/knlZrff4l7dIi8hNmuGlf+yZM27Ha96BWTSt22WtByuZwwmp1L/SISlNd2LaUzof7JyiXMdtUTFUpe6XkIj4YstHyJx+PCNE1heHzoMoiV5odgO9G360fff/X4pLryMVQ0L8D6RVCmhdpAN7VvPOnpCJ9O9srKilNeZIWoNwTU8mQYhuextluP37545alct3ojpFOZkfTfhNmuWpbkRQ/NczTo9iXLspxGDJr55FX+SqWSq0HllddSp7hex+SVT+qxJRIJJ98yyg0afT1+aZZXGpXqX7lcds4ddTsyPxKJhKsOkevQ06ldmRLNY0k0L+Ry23qcTuT5Jtehn0fCp/zK4w8CzXKklgk9/6Qg++OVF3776BWu1gN62pebH5tW63OZf0F4bctrf9uVnXbClCsv8trb7s/vPND/TNMUqVTKt7HXiTx+eX2ET33jl0515UPdss6TDUyz2eD2+60eBuVaFOQaJMuoHt6J37bVMil8yrhfmdF/55eWYcp2u/V0Io9RdpTgU39JKysrrrrBa3tex+2Vlu3ON6G01dTreC/qUblNv/QUzbaE37kSNM28jk2tM2Qey9/rdWeQ+sMvXf3CvbQ71n7W2qLdJkrNO/+brdvtysZPt3LKE8Ywtmq76xXXnqIENTMz01JJbAeyolMrNyLanto1Bmj3kg136i+yM7Ad2w5byWzztL7f0qzdsfaznr6T1yu2bePVV1/d9LGz69nu7Oxs19/DqVar+Pzzzz3fqetkq7a7HsV1fKfkpZdeavk8BREREREFs5522E6zm45Vt+06edVqFZcuXcLU1FTbmXt6rZvtypegbdvGZ5995vnCeCcLCwv45z//GWrmy63abi/Il6zDzIIpj7dSqWBsbCxw/mymXr8oTEQbQ5+AgEhi2ehP6538pl+oE/e88cYbbSeR2elpFuZY+5r+aI+CUV80DvOO0npt1Xa3ihwXL8ftd3q/YyvUfSZRIKLtx2+SCNrdVnwmxqCdT7YjUiEnaOonsnwbhiHiAeae2MlpFvZY+9m2e5K3U/x//9//B8MwAACvvvrqpg153KrtbpV7770XaH4m4J133tl2T/EqlQpuvfVWZxrgY8eO9eyDpUTUW8lkEseOHQOaU3ffeuut22raedoak5OT2Lt3r/PvvXv3hvr8Dm1fkUgEc3NzQPMTKfITS7vNTTfd5HzI/bHHHms7mmqnp1mYY+13EfGfl9CJiIiIiIioD/BJHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1EXbyiIiIiIiI+gg7eURERERERH2EnTwiIiIiIqI+wk4eERERERFRH2Enj4iIiIiIqI+wk0dERERERNRH2MkjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5O1QlUoF2WwWkUhEX7SlbNvG7OwsBgcHUalU9MUbZnJyEoODg4hEIhgYGMDk5KQehXa4hYUFDA8PIxKJIBKJIJvNwrZtPRptgs0+zxcWFpBMJpFMJvVFO0atVsPExAQGBgZcaVapVDAwMICJiQlX/M1SrVa35bXEj5qOtP1s17ZJL9RqNUxOTracw362+tzuRJ5L1L/YyduBKpUKrl+/junpaX3Rlrt06RIuXrwIy7L0RRsmm83iiy++wOrqKizLwtjYGMbHx1GtVvWotEMVi0U8//zzWFxcRL1eRz6fx/T0NObn5/WotAnm5+c37TxfWFjAm2++iaWlJX3RjlGr1TA/P4+5uTk0Gg198ZapVqv4+9//vi2vJV5kOk5PT2+rdKT/2M5tk/WybRvz8/O4cOFCX5S9arWKS5cu4ezZs/oi6ieCdizDMETQLMzn83pQT5TLZVEul11h+XxeAGgJ3wgrKysCgCiVSvoi6iOGYYhcLqcH72j1el3MzMzowTvGZp7n5XJZABCJREJftKOkUqlNS7MwAAS+lmw2r2tXIpHo+f56XcuoO2HaJjuNLHvbqayELbsrKysiHo+Ler2uL6I+wyd5O9i+ffv0IE+2bePChQt6cE+cOXNGD9pU//znPwEAN998s76I+kS1WkWj0cAtt9yiL9rRLl26hB9++EEPpj5277336kHUxkZeu3RbfS3rJ0HbJtQbYcqubds4dOgQ/vjHPyIajeqLqc+wk9fnbNtGMpnckOEF2Wx2y4dQsZHc/3766Sc9aMcrFosYHx/Xg4moaSOvXbrtcC0j6kbYsnvp0iWsra0hnU7ri6gP9byTZ9u2axKMwcFBz0kwqtUq0um0M4lCOp1ueZFVf8F6YmLCNeGC+uJ/sVjEwMAAhoeHnckYbNtGNpvFwMAAIpEIhoeHW7ZRqVRc+5rNZl37WywWkUwmMTk5iWq1imQy6cRdWFhwrQs+2ywWi644+nEtLCw4+5BMJj0nk9DTK8ikE7VaDffddx+Wl5cBwPmtmgZB9tdLMpl0xt3v378fkUjEN5/VNPN6T04/tnQ6jVqtpkdzmZycRCQScRrKch/UiRk6HVulUnG2qaaLHi7p+aaWx3aCnhOd9hcd9kE/J4KkvTqZiXxBXC1X8iV6WS7l+uR+dTp/2tG3nc1mXfku82H//v0AgPHx8ZY88VKr1VzlaXBwELOzs3q0Fu3SVgpaViuVSks8mf6Tk5M4fPgwoByTWm5lHvpto5t8rlQqrklr9D/5e31f1PCg+apvS6939f33qrvbUfO3Xd52Kl/q8avHpobL/db3OUia6+dzMpnEF198oUdzXWMkvSwGuUbo5Ub9U/O0Hdu2ne161QeS13mg5rGsn/W01MP9BLl2SZ3SplN5TAa8lqlmZ2dd9bQ8flln6furh0t6Prerz/3OE69ri1yubzfs/nRKW3iUBb+2iV+adaKX617Uh17089Xr+qtT903fLygTRnmVJ71+8jvXisViSz0m44Utu3bzyfjY2Ji+yKGXJ7/61Yt+rul/vRLkGq/WqzJvZRp2yte+oo/fXI96vS7i8bjIZDLOWF/5DoL6Pk2pVBIAnPdRLMsS8XhcABCFQsFZVz6fd8Z25/N5kclkhGEYwjAMMTMz4/wmn8+LVColTNMUAMTKyoqzL3I8/8rKimu53K5hGM5Y5nK5LEzTdH6Tz+edbWQyGWd98pj0cdlyffL46/W6yGQyzu+9jqtQKIh8Pi/K5bITV38HYWVlxbVeNb2CZKHf+wtB9rcdv3dyZPjMzIzIZDKiXC6LmZkZAY93amS+yHfqZNkwTTPQeHG/fQh6bPV63XOMfb1ed8qL/He78ugn6DkRZH877YN6TgRJ+0KhIOLxuHM+5HI5AUCkUikhmnkjwxKJhEilUs4+pVKpjudPO7lcruW38jjk/kjl5vtYQdYr802+byDTX89fXae0FSHKaqFQcB2bfG/UMAxhWZYQbY5J5pWsBwuFQkvehc1nfVsyrdXtCOV49N/L9er76nXuyWOVZVbWXYZhOGnUqe72Io8hHo+LeDwucrmcUxZlOqiCli9ZvvVj09+dC5vmstzF43FhWZbrfFbXWyqVnG3JfdDLYpBrhJ4Xcr8AOGWuHRlXnuMy/WSYKsj1W2jlTN1fWZeo54wfv2uXDO+UNkHKo/BIv3ZKpZIwTdNJV3mOyt/We3RNCXKeqGVMKPku67/17E+ntBUh2iad0szPRtSHXoJcfyWZlrIc5/N5Jx3Va0GhUGg5t6VO114p02x3euWvFKbsyvTT90fKZDLOuamWG71+9VIOeJ1ZL1luZRmve1zj1XpVXitKpZIolUqe538/a60910FWEGriyYyXJ0q9XndOJpVlWQLNhpD6e1nIZMGxLMsp8LJwy5O4Xq87J00+n285ueXFSZ5IsnLQ46gngDwp1EpTKCeluo1EIiFM03T+LckKQD0J5XHpE4bo6xTN3+thMr3gcQHUyW3pwuyvF7/KRYbreey1H4lEoqXCkWkbpGLx24cwx9ZuHV77C5/y6CXIOSFC7m+7fZDHok9Soh+LPA/1Y5aVpdoQhtYwXWneRAly/niRx69X/F4XcOFx8WjHK64874P8vl3aBimrMl31eIlEQhhKB8NrP2U8NZ+EkieqoPksmr83tBsR8vf6ORom/b3OG68wmUZe8eS21Lrbi9wHtREllLxVz7Ew5aubYwuS5rKxrjckvM5nv32Q6+10jajX6wIeDUT5ez0dvKB5jqt5IBu+6v6GvX7rN0JWQk744JW2arh+bHraeOVju/KohvlJpVItZSCXywVan9fxtKtz2p0ncpl+/fHqkHWzP53KnQjRNgmSZl689m+99aGXRBfXX719IvdLTXevczvotVf+Vo0nf6vuq1/eepFlX89boXTK1HNT1i369cNLmOvMenilqdc1XsZLJBKuYwqTXv2gp8M1L1y4gHQ67XqZc2RkBEIITE1NAQCuXr2KRqOBhx9+WPklEIvFkMlk0Gg0PKdFP378uBMvFou5lp06dQoAEI1GMTQ0BDT35Te/+Y0r3s9+9jMAwNzcHADglltuwfLysuvR7cGDB10TPNx+++0AgMcee8x1XH/4wx8AAEtLS7BtG7VaDUtLS0ilUk4cSb4U+9prr+mLcPDgQT3INb56YWEBlmXhyJEjrjixWAymabrCwuh2f8N4/PHH9SCgObxD/ndpaQmJRMK1/I477gAAvPvuu67woDbj2NqVR1WQc6Lb/W23D7/85S9d/5bk8Bh5jo2MjLiWyzKlj/FPJBJO3KGhIUSj0UDnj5e3334bAPDQQw+5wtPpNEzTxNLSUuAhNrqbbroJhmHgnnvuccK6mZRHT9ugZXV+fh6NRgOPPvqoK97i4iLW1tac+slPLBZDPB53hbV7Ob5TPgPA8vJyy0QIDzzwANAse710xx13wDAM3HbbbU5Yu/LgVXe3Mzg46EqPgwcPIpVKodFo4JNPPgE2uHwhQJrbto3p6emW8x5A22FSfjpdI77++mvAY1KXBx98EADw3XffucLbUfMgFovh/PnzAIAPPvgA6OL6ffDgQeRyOViWhdOnT+PQoUOYm5trSZdupT3eK/r000+d/w9bHoMYGBjA9PS0qwz99re/dcXphl7nqLzOk9dffx3xeLwl7p/+9Cegee1Zj07lLkzbpNs024j6UNft9ffuu+92/fvixYsAgMuXL7vCdUGvva+99hpM03TFi0ajWFtbw+rqqhMWxueffw74XBNfe+01jI2NudI3Go3CMAw0Go22aYhNvM6EvcY/+OCDnmXm+vXrelBf6lknr1KpoNFo4Oc//7m+yOW9994DfDJFVl7dNu6lWq2GRqOBY8eOucYD79271xXnoYceQjwex+HDh5FMJp1CfPToUWVt3qLRqNPg+/rrr3H16lXA5wJy1113AUrnMoz3338fUDqbqsHBQT0osI3a3yDkRBqycbJ3715XPh07dgwAuq7ItvLYVEHPia3Y32pzxkp9zLzczrfffqv/pEW354/sFHpVvPKCrncygxoaGsLa2hoOHjwIu/kupN4I6UbQsirrLr3hFdTU1BQ+++wzQHmvr9u0UPmdS3v27NGD1iWdTmNtbQ2xWAy1Wg3ZbHbdjc1OZOfmyy+/BDa4fAUhz2ev894rrFf8zlmveiUo2ZCVjcNurt9/+MMfYJompqen8cILL3R9bgSlTtSyEeVR1m979+513vMcGhpqabRvpGq1CsuyPMt4LBYL3DBfjzBtk27TbKPqQ1Wvrr+y/Mv3SP0EvfbOzc21pONGWl5extzcXMt+yfPpxo0b+k9abMZ1ZqOu8f2qZ528oP7973/rQQ79zki3vv/+ewBAqVRCc0hqy18sFkM0GsXi4iJyuRyWlpawf/9+pNNpz5dfO2k3y2OQO9R+enkHRLVR+xuGvMNcr9db8kcI4VthdLIdji2MrdjfWq2GeDzekubyTz5lbKfb86fdbHn604hu2M3JI+68804g5PTSfjaqrHqpNCezOXPmDJ588smWp4dhpVIpWJblqkt+/PFHAMAjjzyixOyNWq2GZDKJsbExHDhwAJlMRo+yoTa6fHUS5slZL4yMjMAwDFy5csUVLusV/YnmenRz/Y5Go3jhhRcAAC+99FLH+qHXel0eh4aGsLy8jEQigenpaZim2Xayi43QacZh/YnKRgjTNllPmvW6PtRt9vW3F9fejWBZFnK5XMv+yD+vp+aqzbzObMQ1vl/1vJP34Ycf6kGevvrqKz3Isd47fTfddBMA4OOPP9YXtYhGozh79iwsy0IqlcLc3FzgmcgkdSiI1+xpkj6EYTvYyv2Vd87kMKte28pjUwU9JzZzf/fs2YPl5eV1N7jWc/60GzInh0GGZTdnAL1y5Qq++eYbHD9+3PNOc1hhy2q3d9CLxSL279+PM2fOYHFx0XPIVFgnT56EYRjOrGy2bePPf/4z4vF4xwt3WNVq1RlCtrq6inQ67XmHfCPoZWYjytd2NTU1BcuyMDExATSPfXp6Grlcbt3XU3hck8Ncv2u1Gt59911kMhlYloW//vWvruUbaaPKYywWw+LiIsrlMuLxOMbHx52030zq0FQvattkq3WTZhtRH/rp1fW3U9ww195eP7Vsx+tGURibdZ3ZqGt8v+pZJ08+dl9aWvK8wyOnN5XvyV27dk2L8b+9/gMHDuiLQhkaGoJhGJibm/M8kbymT43FYigWi0ilUlheXg7USLNtG6ZpIhaLOXeXvE4SuQ/dvI/xi1/8AlDeieiVjdrfMO6//34AwJtvvqkvAnzyKYjtcGwIcU5sxf6Ojo4CyvsBurBpH+b8ke8+yA/Zq+Rd1W7vRJ8+fRrLy8u4ePGi51CmbgUtq/I9KPlemMq27bbpWqvVcPjwYaRSqZ5eFIeGhnD+/Hns2bMHt956K+68804MDw9jcXFRj7puhw4dApqdjs2il5mNLF9ByHdFOr2b00vpdBqZTAaff/45IpEIDh06hBMnTuDs2bN61FD0a3I31+9sNot8Po8XX3wR8Xgc586da1s/9NJGl8eRkREsLi7CNE2cO3dOX7xh5NPbRqPheTNDbZtslG7bJkHTbKPqQ12vrr9B4wa99sr98rpmVJqfighL5pnX0MuxsTEsLy97rrdWq3U8ZzfrOrNR1/h+1bNOHgBnGEQqlXJ1rorFolNwHn30UacDpjd833//fZim6XlCe3XWJHlxUaXTaViWhWQy6dpOtVp13iuYbH77TvXcc8+5/i3pj/RrtRqWl5fxzDPPAM0Cnkgk0Gg0Wk5KOeZbvhAdhnyRWH9pGR3SpJ1aczx8r/dXz89OhoaGYJom5ubmWr6tUywWux721O2xqeXItm1nCJ7XcQVN+yDnRLf7G3QfvMghXMeOHWv5vkw2m225M+a1rTDnj0rGeemll/RFmJubQyaT6bpx4pVXXvVDJ/rxBi2rcsKV6elpV17azbuPfp0L27ad9/50+r6EJctasViEEAJra2uYmpryvUDq25NPbfze+VJZlqUHtdSdqm7yRmU3JzlRn1h1U770fZR39b0aQ53cd999MAyjZVIieGynV7LZLA4cOIDFxUVn+LCcyGM9Xn31Vded+LDX72w2iyNHjmCoOVmTnJziyJEjLeUsCH2bnYQtj1K77ejnfzQaddoBul5dU7zOkxMnTgDKzUJJtk28hrCF2Z9OwrRNwqSZJF+70enrXq9ur7+6S5cuwTCMjnGDXnvle2bZbNaVvtVqFb///e99h5K2y8v//u//BnyGlMshlaOjo64Ond38xpx8PxHNTqa+nTDXmWq12lJmgtK3C5/zg5r06TbXQ07vKqdcldPS6lMmq58lkNP/zszMtEwra1mWM4WtOjWqaG5LfgfD65s7deXbGXIaVTnFqzoFsWEYznSycp3qtPByGlbDMJzpmq3mt2D06ePVb/DI45DT0qrT7daV79OoU9nKb/pAm8paToks96Hc/G6NXEcul2s7/bg6pXKhUHCOI+j++pH5mEqlXFPnq98nkdRjVtct96FTmfGi5rE+fXiYY5PpHo/HRalUEoXm94HkVMny+NqVRz9Bz4mg+9tpH8Kkvcw/NKc4TzSn+VenOy43yz88phwOcv74kWUyo32XSE617hXXa5lOTo8cj8dFuVx25WMikRAzMzMtU45LndI2aFmV+yD3Q25fne5dTjNumqZzPqtlZWZmRpSa3/qR+1QoFJz8C5PPpmkKo/lZB/kn6xE9LeR6883vY+VyOTHT/DaTXIc8VhlX/xSIDJPHJc/RXC7n/L5d3e1FrRtnZmZEXfkml57+ImT5kuddoVAQpVJJ5HI5Z6rxeDzeVZqr55ZMy7zy3VVZp4g23zaT6+10jZBlKd78jqv8K5VKolwutxyvF1nuMpmMsJTv+hke360Mev2WYfr2ZRp45ZvO69oVJm2ClEfR5lrmRbYl1PQ3TdNVLnpxTel0ntSV658se/Kc8LoeBtmfMGkrQrRNgqSZrr5B9aGXoNdfoWxLPR55Dqt1vPCZ7l8EvPYKpfzCpw2rritI2RXNc10vH5LMT3l8XtcuuT1T++RE0OuMrK+gff5DbqvT/ucDXuNlniS0Tyio59Ru0NNOnmhmoExEKBdZXalUcioomeBqwZUnh/qXUL7Hoi+DR39VXqj8tiELh6xIvPZX7keq2XhFs0LT40krKyuu45eVquR3XLLgqn9qYc8rHylNJBJOxZnJZFzH5EVWqPIEVHXa307kySXTQz8GNDsHepi6H/JY0CFtVV7r9Fpv0GOTDRLDMJwLhUzfcrnsuT21PLYT9JzotL+d9kFfBp+0V39Tan6kFs1KWy9z+m/15Z3On3byykdk/fJd376+/7q68gFXs/nhctlYMJsdKi+d0kkKWlZnZmZc6epV7mT6yvNZNC+gMj3lumW8XC4X6hyT+y87LH5/6gXcsizn+GQjptz8yL3sXAmffBHN9JHHLY9L7lsikXBd4PXfdmJZlsg0Ox7wKK+6IOVLNPNexks1G9OybMu00fcXHdJceJxbsgGWSCSc8iDTOsh6252P6rVU/zM8Omo6ecxB0ksEuH6r+6qmidfxtqNfu8KkTblcDlQeJblv7Y5bNBuJ6nF4XVdFD64p+jJ4pFVd6YyjwznR7f54pa26jXyAtknQNNNtRH3op9P1V1VofsxcxtXLvyT3w+tY9frBK069Xhe5XK4lfXVBy65Q8suPmp9eabDS7BBntA5p0OtMvXlzIq7d5InH48653u4Y6gGu8WpZk39eZQIe51S/iYj/VCbko1KpYP/+/cjn8z0Z/kJEtJkqlQquX7+OBx54ADdu3HAN1fniiy9w5coVrK2tuX5DO8vk5CTuuece3Hzzzfjqq69cQxIvX74M0zRbhqIR0cbaju1H27Zx33334cyZM56vRnWrV9cZ+R6f1zBPCo+dvA6240lKRBREtVrF+Ph425ffk8lk2+W0vRWLRfzrX//ynWTFtm08++yz7OQRbbLt2n6sVqt4+umne9aZ6tV1plNdRuH1dOKVfsQXOolopzp//nzbiQqq1aozIyjtTNlsVg9ymZ+fDzQhEhFtjF5/22+9hoaGcPHiRTzxxBNtrw9B9eI6IyehYQevt/gkr4N0Oo25uTmkUineCSWiHWViYgLnzp2DaZp45plnnCF9N27cwEcffQRs4PTytDmGh4edD00/+OCDeOCBB4DmrKjvvvsujhw50tNhWUTkrdb80Pm+ffswNTWF+fl5fPvtt9u2jq3Vapifn1/3U0ZeZ7YvdvLaiEQiehCYXES0kxSLRbzxxhuuD+umUik8+eSTG/pxYdoctm3j0qVLuHz5MpaXl4Hmh43T6TSOHj3qO9U6EfWWbdt44oknsLS0BMMwkMlkds2TKV5ntid28oiIiIiIiPoI38kjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1EXbyiIiIiIiI+gg7eURERERERH2EnTwiIiIiIqI+wk4eERERERFRH2Enj4iIiIiIqI+wk0dERERERNRH2MkjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ28DVapVJDNZhGJRPRF28bs7CwikQgWFhacsIWFBSSTSSSTSVfcXqvVapiYmMDAwAAqlYq+2CWdTmNgYAC1Wk1fRFtkdna2Y75tlGKxiGQyicnJSX3RlrBtG4ODgxgeHtYX9ZRt25idncXg4GBXae93zlUqFQwMDGBiYsIVf6fyqteItpp6/m6Xumsn2o711Xrr5rBs20Y2m8XAwAAikQgGBwdD1Xeyfaq38/zCNxPr795gJ28DVSoVXL9+HdPT0/qibW1hYQFvvvkmlpaW9EU9VavVMD8/j7m5OTQaDX0x7QBHjx7FjRs31nWhnZycRCQSCfw3OTmJYrGI9957b8PL6HY0Pz+PixcvwrIsfVFHPOeIttalS5fw7rvvdnX+0vZ26dKlruvmsGzbRjKZxNDQENbW1lAul2EYBn73u9/pUT1Vq1V88MEHLe1Tv3DaoQRtOMMwxE5L6nK5LACIRCKhL+q5VColAIhyuawvoh0ilUqJfD6vBweSz+dFPB4XlmW5wgC41mlZlojH406YLKPdbncnk+nT7TnDc44ouF7XMTMzM4HrriBxaGuUy+WWOnS9dXNQsgzV63V9USh+7Ty/cNpZ+CRvE+zbt08PIsW9996rB9EO87e//Q0XLlzoeojK3NwcYrGYHuwSi8UwNzenB1MXeM4RBVOpVPDhhx/qwety991360GeNmLb1DtnzpzRgzbNtWvXAADRaFRfRORgJ4+I1i0ajSKdTuPIkSP6oo7uuOOOjh08KRaL4ZZbbtGDiYh6rlqt4te//rUevCm2ctvUWTab3dLXBf7973/rQUQtet7J018EHR4e9ry7X6lUMDw83PK+jfzrNm4Q+vq89rFWqyGZTCISiWBgYADpdBrZbNYVx0u1WkU6nXbWnc1mYdu2Hg1ovvsm90O+QKzGtW0bExMTzrr8JkKpVCqubabTaVSrVWe5OkFFtVrF4OAgBgYGUK1WA70oXKvVnPUPDg5idnZWjwI0360aHBxsG08vH8lkEl988YUerYU6WYRKPTa5bpmexWLRFVcl81b9U6nharroeZbNZl0Twai/U1+q91qfnvbFYhEDAwMYHh72LTMIkN9ovrQs80LG0SesCVLG9fzyOlekAwcOwLKs0C9Kp9NpPaito0eP6kGB8l4/N73SRKefOzK9vF5wr3i8rK7me6RZ3vVwffIFvYzp9UJYeh76nXNeE9nUajVMTk46551adw4PDzvlbmFhwSlvyWTSc3871Q9hzuVOZVc/t3R6GuvnMULuT9C6WlLzP9Kse9rVSeoyNX/049DLil5+1bpf6pQvforFYksaBsl3mZ6Snlf6eSb3tVKptKSPmrd+4d2m0eTkJEZHR9FoNLC0tOSsW6Ufm1/a6fE++OADPYrLwsJCx23rx+VVhv10uobIiddkXslrr0w/BKxzJbX8queFX7lG8xom66zh4WGk02lXvvrVV2o7IUi9pNZpXn9eksmk887a/v37PfcfzWuOV1lW6fmol0+dXJ/sYHqln17nDw8Pt82fMDqtWz9PZX7r6aqWAzW8Uqm01AnqutVr7GTz2hSJRDznBVDLrr79iE+e9R19/OZ61Ot11zszKysrwjRNAUCsrKw48fR3acrlsvPeWqFQcOKFjRvEysqKACAymYzzb8MwhGEYrrHN6nFYliUSiUTH8clyXZlMRtTrdecdIgBCT+pCoSDi8biTLrlcTgAQqVTKiZPL5UQqlRL1el3U63UnjqpQKAjDMJzx3/L4DMMQlmWJUqnkvH+Ty+Wc4wAgZmZmxMzMjLOP6hhyme7xeFzE43GRy+VEJpNxjmVmZkbZCyEymYxrX9VtSLJ8yPev6vW6a51+Y9jr9brI5/Mt7zbqx5bJZESpVBKlUskzT1Xqtk3TbIlXr9cFtHKby+VcaS3LomEYLfHg8b6F/h6Umvb5fF6kUinP80XVKb+FMlZfnh+FQkHAY3x9pzIe9HyWZJmR59Z6yPca9DRUye0FyXu576VSSYhm2fHLeynffFdQHpNMC5mPal6urKw4+a6noVrGJVl25P5IQeoFEeK9j6DnnHouyTTXz7tCoSAymYwol8vO9uPxuGe4vr+d6oew53KnsutXr4mA53HY/QlSV6vkNqGdS/Lc9SqX8XjcdW51Kiv6Meh1vwiQL37k+aDXOfF43BVPLXt6PHl8al7NzMw4ZUnGVfNVTTf93BHN7antgl6kkb4PUtC0y2QyrvImjwsd6jfRZttByrCfTtcQPU1SqZQoFAqiVCo5+VQoFDzD221b1rn68cj0UNOiVCoJ0zSdciOvYWHrq3w+L8rlslPn6eldXkf70q8OluGdyrIIUD7bkeVNZ1mWMJR2qFrnq/WH5LVffuFB1+1Xj8k804+v3mxvyTT3qr/1a2wulxOFQkGUy+WWuFIikXDKkbqvpmm64vWz1hKyDvl8vqVQyBNbzdR4PC4Mw3DFkyeGXgjDxA3C68SUBUev9NTCKSvxdkzTbIljWZaA1sCr1+uuSlbSK0rDoxGoXkTlevSKK5FIuCp7WZEZSsNE3bZXmsjf6CepzE99Xeq/hXLSqnknL3ZqPNFMN337XrwqNbmfiUTCtV6vY9LJ9NPLl1AuIpLcjl7xy4uPmu8yrp4vXvskw+Tv6/W674UyaH57pZMsW1KQMh70fFbped4tmS76sapkOqsNRuGTzolEomVd8rzXG2Uqmb/6NtSLjcorTCjpL7clG0iqoPWC8DlGL2HOOb9yK8uTvr9qg08l1y0FrR+CnstByq7w+J3o8jzutD8iQF3tRZYh/XemaXqeQ2oZDFpW5DHo9bX8b5B80cl1qtuW+6M2nmQ6yYa65NXglnFzuZwrrlddJn+vp5teBnqRRsLnnA6adnrnRPJKAy9+20bAMqyTaaJvN+HTZtDrPVn/6+Fy2/p6dV77J7el/jaVSrWUhVwu50pHr98Jpczo5cNr2+tpX3rVA2q4/nu9LActn3709anhXp0Yrzpf+KSLX3iYdcvjUMuJbA/r7Qd5o07ll77QbmoLn7Igy6p6fZdpDo96qV/1dLjmhQsX8Jvf/MYV9rOf/QxoTqwgLS8vt0xG8sADDwDNx+2qMHGDuOOOO2AYBm677TYnTH/H56abbgIAnD592gmLRqMtx6ZaWFiAZVkt7yTFYjGYpukKm5+fBwCMjIy4wmU8+Rh+YGAAf/7zn12P7p9++mnn/+fn59FoNPDoo486YQCwuLiItbU1DA0NucJPnDjhvKSrb9vP4OCg68XegwcPIpVKodFo4JNPPgEAvPbaaxgbG3PFi0ajMAwDjUbDefw+PT2NdDrd8qLw2NiY69/dePDBB1vWCwDXr1/XgxzRaBQnTpxAo9FoGcrw5ptv4sknn3T+/fbbbwMAHnroISXWf4YamqaJpaUlz6EYQZ06dQpo7pOeb1LQ/I7FYojH4644etoEKeNBz2fdZk/N/9hjj7UcHwB89dVXQHPIzNLSEhKJhGv5HXfcAQB49913XeGq22+/HfDYxh/+8Aegea62G1ojzc3NwTAMnDx5EtlsFqZptgxTDVovBNXrc07fX7lOPXxwcND17yD1g6rTuRyk7Prp5jzutD8IUFd7+e1vfwsAeP/9913hY2NjLXWSHFIm9yNsWfGq+8Pmi/Taa6/BNE3XtqPRKNbW1rC6uuqEvf7664jH4y3v2/7pT38CmvWL7pe//KUeBDSPX5K/f/XVV5UY/0kT9frbizTyEzTtXnnllZa0QnNoe7e6KcNS0GuIpNd7N998s2e4rCd7ZWBgANPT067jkOdLUAcPHtSDWvK81+1L1eOPP64HAc3rEboon0HUajUsLS0hlUrpi5xJYl577TV9USBh1y3rv6tXrzphsj08NzfnSl/93O1k3759LfUKANckRV9++SWgTXIUjUad/P7++++d8H7Ws05erVZDo9HAsWPHXGNe9+7d64ojqRcD1Z49e/SgUHE7SafTWFtbQywWQ61WQzabbbnYxGIxZDIZTE9PY3Bw0LnYer0LJMkLtVdlpzd6qtUqGo1Gy/hg2XD+9ttvgeaJs7y8jDvvvNMZG6/ug2ycehX2jSRn5pMn0fLyMubm5lqORzb2b9y44ZzoP//5z5U1/YdX2GZ56qmnYBiG08lCs5xeuXLFdZGQZcCrsSc7VN1UymEEze+pqSl89tlngPLuhb5vncp42PNZ59dA3Ew//PADAODrr78GAOzdu9d1LMeOHQPa1C/tRKNRp9Mo199OLBbD+fPnncb72bNn9SiB64Wgtss5F6R+CKNT2W1no87jTnW1l6GhIcTjcUxPT7s6h2trawCA9957zwl7++23XevrRVnpNl/m5uZarmm6arUKy7I80zkWi7V0hsKIxWJIJBJYWlpy1UEXL1503XDoRRr5CZJ2tVoNy8vLnmnl1U4Iaj1lOOg1ZKvJsr53717nXcOhoaGWDlEv+NX/3bQvg/jpp5+ADSqfss7XH1wAwF133QV0uEHbTth1yxsJr7zyihMm33mF1vm7fPlyy83CXvGrx+TNwn7Xs06e7BWXSiU0h4G2/MmKJZVKwbIsVwX9448/AgAeeeQRJwwh4wZVa760PzY2hgMHDiCTyehRMDU1hZmZGaytreHw4cOuSQa8tGvw6mq1GuLxeEv6yL+pqSmg2SEtl8sYGBjA+Pg47rzzzpbJHrYDy7KQy+VajkP+pdNpfPfdd/rPtoVoc1ZIy7Kci+f8/HxLmZAXby/bcTr6SqWCwcFBnDlzBk8++WTLUyx0KONhzuftTpa9er3ecgxCCN+LfK8dPXrUadx6TdAQtF4Iarucc0Hqh7Dald12Nuo87rau1u92F4tFPPnkk8hkMs7dbtu28dlnn7mesvSirGxEvkiyIetHf3oSlrzr//LLLwPN+k5/Ot2LNPITJO026knBRpXh7WRoaAjLy8tIJBKYnp6GaZobMknGRrQvg9qI8ilvbHrRn9KGFXbd0WgUqVQKy8vLTvrOzs7irbfegmmaeOmllwBllEKvyTbPRx995Aq3bRumaXrucz/qWSdP9oo//vhjfVGLkydPwjAMZzYu27bx5z//GfF4vOXCEiZuENVq1RlCsrq6inQ67XlnAs1G2TfffIN8Po/l5WWMjo6G6sz52bNnD5aXlwMN8xoZGcHq6ioKhQIA4Fe/+lVL46Gbu6G9IIe7GYaBK1eu6It3DDn8Rz7Ne/311/HUU09psf6jXUNSpsdG65TfxWIR+/fvx5kzZ7C4uOg5bEXyK+NhzmcvG3HHtVvy/JbDi3tNHfrdzsTEBE6cOOEM29TrkjD1wk6yUfWDX9kNYiPO4yB1tU7e7X7jjTec/x48eBAPP/ww0LzhdPXq1Zahn70oK+vJF78nRbpPP/1UD3IJeu7o0uk0DMNAsViEbdt4++23W+rsXqSRn/WkXa+spwx3uoZsB7FYDIuLiyiXy4jH4xgfH/ecQXE9et2+DGMjy6fX7MmS/vpQWGHWLV95mZ+fd+pm2fmzLAuVSqVllEKvDA0NOSM+ZD1cLBaxvLzcMtS7n/Wskzc0NATDMDA3N+dZaNX3C4aGhnD+/Hns2bMHt956K+68804MDw9jcXHR9ZuwcYM4dOgQ0LwTHEQ0GsXx48cxMzODRqPhjKPW/eIXvwCAjlMjA8Do6CigjMnW6e+HoXlRkx+/lAX0wQcfBJQx+irbtj3X0wvyjo68Gzs2Nobl5WXPi06tVkOlUsE999wDNB/Lbzdy+JdlWchmszAMo+UplRyH/s9//tMVDo/02ChB8rtWq+Hw4cNIpVKBL1JeZTzM+awzDEMP2lL3338/0HzP0ku7Y2nHbt4R1MuKl4WFBTQaDRw/fhxTU1NoNBquKf/RZb3QznY554LUD93yKrvtbMZ57FVX+5ENnqWlJczOzjrn+MGDB2EYBl5//XW88cYbLe9f9aKsdJsv8g651zYqzU8gjIyMOE+tvdYf5tzxI9+nvnTpEmq1Wsu6epFGfoKknRzCthTwvd2g1lOGg1xDtpuRkREsLi7CNE2cO3dOX7wuvW5fhrER5VOem143IGQZ1J94B9XNutV67NKlS857ivL9ytdee61llEIvvfjii0ilUnj++ecRiUTwyiuvoFQqtb3x3W961slD8+JmWRaSyaTrjmq1WnW9X1AsFlGtVlEsFiGEwNraGqampjzHmIeJG4RlWXpQy2PoSqXSMpyq050G+cK4/rIwlBNAkhfsY8eOtWwnm8064/X1RvrQ0JBr2J28Czw9Pe2qEGzbRjKZbKno9ePsht2czCGXyzkXVTmsYXR01NUwsJvfU7nrrrtw3333wTAMLC8vt1Revdiv9ZL5Oz09jT/+8Y/6Yjz33HMA4AwxUM3NzSGTybQ0MvTjknfAvMaIyyEi7QTJb78hQnoZDFLGg57Pajg8KvpKpRL4CctGGBoacl721r/lVSwWAw1r1POy1nzf5plnnnGFe6lWq3j11Vfx4osvAs10VRv2UtB6Iahen3N6GQoqSP0QRpCy66eb8ziITnV1O/Ju98mTJ11Po+SNpz179rTsU9iy4pXf3eaLHCqZzWZd17pqtYrf//73ToPtxIkTQHOIlkqeO3Kyhm7J+nB8fNxz0oZepJGXWq0WKO3U93bVSYIQsL73UqvV1lWGg1xDVEHTJAy9HpETZKnvoOn1dDQaDVTXhtXL9mXYa1zY8hmErHcaHpPJySHhcuRSWN2uW9Zjn3/+uTPCZ6j5PvLc3FzLKIVekWX63LlzWF1dhRACn3322a7q4AFe86+uQ735TSY0PxmQSCScKWrV6WDlFNH5fN75k9+70Kc1DRq31Px+kfwmhh857Wym+f2STPN7P2hO35xvfldFTr0qp3/NN7+/0m7dclpkwzCcfcxkMs70srlczkkHOeUwmp8pkFMYq9Puyv2U25T7pU4NLKeZRXNaY3l86vTKMo5f2shvl6jbltOUq+lgNb/7p0+fLJRjb7cf6jHLdM4r3yJLpVIt0x5L9XrdSUd1ner3Z9R98vseSzsyD/Rjk+QxZrRvxJge37Qymp9mKDS/I5TL5Zwp0+PxuJOm6n7q6/DSKb/VKYJnZmacT0GoaTczMxOojAc9nyU5ZbGahzLPvaZd9qPmtVdZk8Lk/UrzG5Zonp9yKuh26xfa9OoyjdXzQKWeMzINZVx1amehpJWadyJgvSB8zlkvYc45mZ5qmvidd37hlmU54Wo5CFI/BM3PIGVX/Z2eRkHP46D7IwLW1e145bEsT2oaqYKUlU51f5B88SLjyfTxqhfU+kNOYy7PB71elmmqTpuvljG/z5ykUilhtPncQy/SSC3PuVzOiRMk7dR6J5VKiXK5LAqFghM33rwW+Om07U5l2Euna4jwqQvahct16ueKTuazrItyuZyzTqPZzqs3P4URV74fJ+sVtXx47YtaZtRzT62bu22L6mTZknWorOPDlOUg5dOLWq708iOXGR7fUfSKK7evHm+78KDrlvzqMZl/funsVX/LdRlaXS/XpZZ/NX/U/C2Xy86+7wY97eSJZmFWK79UKtXSIJSNXb8/tTAEjSsbTfHmh7v9rCgfdE4kEmJlZcW5ICcSCWFZlig3P64o46lxO5GNDfU3iURCZDKZlt+Xmh/8RLNwykpCMpuNULkPpvIxZ9XMzIxrPWoc9WIs/9QCri9LKN8qsSxLZJrf2fLbR5V67PF43HNf9WOWlWMikfCML5QGk/onf6OHe8VFwHsZM82Pl7aTz+ed/ZeVsddFrVwuO/FSzQ6c3GdZZvV9DLOfehqqCs2P3UJpBOSV71DV6/XAZTzI+SzJRoZKXhQ6pask91P/08td0LxXy7M8F9Eh71RynalUyjkXvX7rtd/6/sjzTg+Hx/d92tUL+m/Rodzo6/M658KkZ5hwdd/b1Q9Bty/DO5Vd/TdqORABzuMw+yNC1NV+1BuAqniHb+3peaumt98x6Nrli59684Pv+rVOJ+uPdtcQfR/lfupheh6K5vG3u96LHqRRSbmBrKdNkLTT6518s6FpmqYoFApt66BO225Xhttpdw3xS5Mw4X6s5vfQZHrJ9pdpmmJGuWmTSqVc65XpJnlt06/MeNXNcl1B25d+5H7ItNd/D5/9ClPf67yOR25HWllZcTpJ8CmbXuvJNztCXuFh1q3zOncty2q52SPp22+Xj35lwWp+uF1fJv/UmwP9LCL+k6CbqlKp4Pr163jggQdw48YN13CpL774AleuXHGmkQ4TF813Xz7++GPPKcqJ2kmn03juuee21aQhO4Vt27jzzjvxj3/8o6/Sr1KpYP/+/cjn8zh+/Li+mIiIdqiw7UvaWbLZrPMeoPpd0x9++AFzc3N44YUXAg/136k2vZNXrVYxPj7e9sXWZDKJxcXFUHHRbGg+8cQTmJqa8h2TTuSlVqthbGxs06bT7zdyeut+6wixk0dE1H/Cti9pZ8lms3j44Yd938FbWFjA//t//6/vO3k9nXgliPPnz7e8eKuqVqvODFBh4larVVy6dIkdPArMbk6ZjOYMV+udCGC3KhaL+OGHH/qyE9TtBAlERLR9hWlf0s5SrVYxPT2tB7u8//77zkRE/WzTn+RNTEzg3LlzME0TzzzzDO655x7cfPPNuHHjhvPRQvl5gzBxicIaGBhAo9FAIpGA3fzgMIUjvz/jd7dsp0un05ibm0MqlWqZUYyIiHYmti/7V635oflGo4FMJoOhoSHcfffdQHPY5uXLl3Hx4sUN+3TDdrLpnTw07/y/8cYbrg+qplIpPPnkky2NxTBxicJIJpNYWlpCJpPBiy++2NW0ydS/IpGIHoQtqC6JiGgDsH3Zv2q1Gl5++WVcuXLF+XSaaZpIpVL4wx/+sGvae1vSySMiIiIiIqKNsenv5BEREREREdHGYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1EXbyiIiIiIiI+gg7eURERERERH2EnTwiIiIiIqI+wk4eERERERFRH2Enj4iIiIiIqI+wk0dERERERNRH2MkjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYyVOk02kMDAygVqvpi7Yl27YxOTm57fe32/2cmJgI/ZvNUqvVMDExgYGBAVQqFX3xjjE7Oxt6/yuVCmZnZ/VgoMMyIiIiItoc7OTtULZt4/Tp03jqqacQi8WQTCYRiUQC/U1OTuqr2zD6fkrVahXDw8OIRCIYGBhANpuFbduu3549exaXLl1CsVh0hW+1Wq2G+fl5zM3NodFo6ItdJicnW9Lf7y+ZTOo/31ATExO4//77MTIyoi9yVCoVRCIRV0dwZGQE999/P9LpdEuetVtGRERERJtjV3bybNv2fNpQLBaxtrbm6oxsR7ZtI5lM4ujRo4hGowCAxcVFWJYF0zQBAOVyGUII11+9Xkcmk9HWtnG89hPNDt6hQ4cwNjaGXC6HgYEBTE9P44knnnD9HgD+8Ic/4NSpU6GfNm2kWCyG48ePIx6P64taHD9+HPV6HalUCgCQz+db8kUIgZmZGf2nGyqbzeKOO+7A0NCQvshh2zaOHDmiBwMAhoaGMDo66pln7ZYRERER0cbblZ28S5cu4YcfftCDd4xnn30Ww8PDLQ30WCyGwcFBV5gqGo1iamoKt9xyi75oQ/jt5/nz5/HJJ5/g7NmzOHv2LD755BMYhoGlpaWW4ZnRaBRvvPEGfv3rX2+7J0P33nuvHuQpGo12jHv06FH85je/0YM3xOzsLK5cuYKjR4/qi1z++te/wjAMPdghf+/1ZLjdMiIiIiLaWLuuk1csFjE+Pq4H7xiVSgVzc3MdG+jtrOe3QbXbz7/97W+uJ3vRaNR5wvjTTz8pMf9jZGQEAwMDOH36tL6or3ilVa/Zto2TJ0/ihRde0Be5VCoVWJaFxx57TF/kcuTIEYyPj7d0zjstIyIiIqKN0/NOnm3byGazGBgYQCQSwfDwsOdQu1qt1hJvdnYWAwMDgMe7TJIarr/DVKlUnPe8vLY9OTmJw4cPAwDGx8dd61An0vAyOTmJwcFBRCIRDA4OYnJy0vVkSQ4BHRwcRKVSQbVadd6TGxwcRLVada2vUqm41pfNZgM99Thz5gwMw2h5OrZexWLR9Y7cxMQE0ul0oH3y0m4/1Q6eJIeaesUHgLGxMUxPTwd6mtepHITNK/kbtbwmk0l88cUXerSeq1QqSKfTzrHIPNHLflCXLl1Co9HA/fffry9y2LaN3//+9zh37py+qMW+ffuA5np1fssmJiYQiUSwsLDgCiciIiKiHhE9VK/XRTweF/l8XgghxMrKijBNUwAQKysrTryVlRVhGIbI5/OiXq+Ler0uMpmMACDUXbIsSxiG4QqTvwcgEolES1gmk3H+bRiGMAxD1Ot1J165XBYAnH0Uzf3O5/Oe2xJCiHg8LuLxuLAsSwghxMzMjAAg4vG4s+6ZmRkRj8cFADEzMyMymYwol8tOXHVf5XGVy2UhmvtkmqZrn7xYltWyLl0ikRAAnHWryuWyZ3gmkxHxeNy1PzItOu2TlyD7qZJ5p5YRXT6fFwBEoVDQF7kEKQdh8koo5VqWAb28eqWpTu6/X3p6hRcKBWEYhnPMlmU5+63vY1DyfGwnl8uJmZkZIZT9bneMAIRpmnqwED7LZBmV2yAiIiKi3mrf2gspn8+3ND5LpZIAIFKplBNmmqbTCFfJRrNKNgh1ekPXqzGay+VawsoenTzJa1tyvbKDJ8lGvroeGTeXy7ni6uuVnQxVqVTy3CeVTEt9/Sq/Tl69XhepVKolvFAoeB5fp05JO0H2UzT3SXZkvMqDSuabWo68BC0HQfNKNPNav1kglA6TnqZe2qVnPp9vCZcdZb1TK9NBP8+CqNfrAs2bE37K5XLH80on08yrk+61zLIsUSqVXPGIiIiIqHd6OlzzwoULLZNH/OxnPwMAzM3NAc1hgZZl4fHHH3fFW6877rgDhmHgtttuc8J6McHI66+/jng83jLj5p/+9Cegecy6X/7yl3oQ0Bx6h+Z+LS8vuz4NcPDgwY77++WXXwIBj2v//v3OEL9IJIJbb73VyQPVqVOnkEgkWo5vPYLu56233orDhw+j0WhgenoayWSy43DMf//733qQS9hy0CmvbNvG9PQ00ul0yzDTsbEx17+DkMOE1T+vd0RffvlloPntxl75+uuvAZ/hslCGaU5NTemLAvF6n1JSl8ViMRw8eNC1nIiIiIh6p2edvFqthkajgWPHjrkasHv37nXF+eijj4DmZBq9lE6nnc8fyPf9vDpgYVSrVViW5dkojsViMAwDjUbD853Ddh566CHE43EcPnwYyWTS+X0vJ97QP6GgTuMv1Wo1WJaFBx980BW+WYQQsCwL+XzemV1Tf39Lt7S0pAe59LocXL16FQDw85//XF/kGdaJ1ycU8vm8Hg1XrlxBIpHQgzfUX//6Vzz99NNdd/ivX7+uBznaLSMiIiKi3upZJ+/7778HAJRKpZZGrPyTDe+NUqvVkEwmMTY2hgMHDqz7m3DtnkxAmVgirGg0isXFReRyOSwtLWH//v0b/vHoaDSK5557zhUm82wryW/O/eMf/wCaT07Xq5fl4LvvvtODeu748eN6ECzL0oM2VKVSweeff97TGw1EREREtDV61sm76aabAAAff/yxvshTrzt71WrVGVa5urqKdDrddpheGJ9++qke5KIODQwqGo3i7NmzsCwLqVQKc3NzXc+YGNTIyIjnE9Rvv/1WD9p0IyMjiMfjWFtb0xe5dHq6tZHlYCN5dfRWV1f1oA1z5swZLC0t+Q4llcN/2z21fuCBB/QgR7tlRERERNRbPevkDQ0NwTAMzM3NeT6Rku+fyaFgnTpOYR06dAgAun6fyMvIyIgzJNNvWn3TNLse3oZmehSLRaRSKSwvL7dtRN9zzz1ADzpltm1jcnLS6ZxeuXJFj7Iu3e5nNBrt+J7bnj179CCXXpcDeSyXL1/WF22IiYkJAIBpmrAsq6c3Q+666y7Ap/N45MgR5PP5lj/Zqc5kMsjn821vaMgbPV7aLSMiIiKi3upZJw/N96Esy0IymXQ1TqvVKt577z0AwMMPPww0J/xQO4NenSiVV1y1seo1vO2HH37Qg1rYtu3ZKZVOnDgBAJidnXWF12o1LC8v48yZM67wICYnJ1uOVx9K6eW+++4DevAU9PTp085kK/F4HJZltRyf1zfgbNtGpVJpm17ocj9t28bq6qpvOnz11VcAgNHRUX2RS7flwM99990HwzBaJsrBOtfrpVgs4o477gAA5/3JXC7niiPTQVepVDqmdzQadfJbl06ncfz48ZY/+b7m448/juPHj3ve0Pj00099v4notaxWq/EbeUREREQbSZ9ucz3k98TQ/BRCIpEQ8XhcGIbhmkJdTqtuGIbI5XLOn/ydSn67TE7/n8/nnSn60ZwCf2VlxVmn/OaZ/PabjCOnqJdT05um6cQTzX2XU+Kr09arxyS/6yW/V6ZP559KpZztSep61W+PGYbhTCMvP2/Qbmp7KZVKCcMw9GAhtO8K6lPyy+W5XM713TI5Jb/Mr3w+L1KplHMs6nrkZyM6fe5AtNlP9RuHclp+mZ7tvpsmy4f+GQNd0HIQNK+E8pkJmR6yHMr1plKptp8EqNfrzn55ffqgXq+LmZkZYRiG8ymLer3u5KX8hmImk3HyQF2P3D/9e3Re5Pnk9bkDL50+oSA/y6B/iqLdMpkW7fKbiIiIiLrX006eaDbs1A9Fp1KplgalGsdofhRdNL99B49+p4xrmqbTAZMdRNkoXlE+vJ5IJMTKyorrm2Lqd+Bkw1WPp/6pnRu5v3qjW6X/Hs2GsR4mO1KJRMJZH5qdkk4dGKF0yvQ0lQ3nIH96o7tUKrWknUwj9ThlZ0L/dpsXv/1UOzxo5r/X9/t0ft9W1AUpB3p6oE1eSWoamabpfNcwkUi07eDJdAzyp3fy1ZsXcpvqsajxgnxrUCidx6AdrE6dPL/vLLZblsvlXDc5iIiIiKi3IuI/HZRtIRKJAPjP1PrkL51OY2BgoGfvnXmZnJzE+Pg48vm856QgQfRqPyuVCn7961/jm2++8fycxW5SqVSwf/9+JBIJLC4u6osDmZ2dxUsvveT5bl5YyWQSDz74oGcZabeMiIiIiDZOT9/Jo83xt7/9DZ999lnLe33bTa/288yZM/jHP/6x6zt4vXL06FGMjY21vIcZlpwkyKsT124ZEREREW0sdvJ2IPmdvfHx8XV3oDZSL/ZzYmICR44c8fz0A3VvamoK1Wq1ZTKZoKrVKl577TW89dZb+qK2y4iIiIho422bTp4+Gye1F41G8dZbb+Gf//xnx1kVuyE/f+A1y2YY69nP2dlZ/Pa3v0U6ndYX7Vo3btwAmjPLdprltBM5jLbdZzu8VKtV/POf/0SxWGx5utpuGRERERFtjm3xTp58/0u1nnfBaH3ku5GqbVBMdr1kMomlpSVXWLlc5lNOIiIiInLZFp08IiIiIiIi6o1tM1yTiIiIiIiI1o+dPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1EXbyiIiIiIiI+gg7eURERERERH2EnTwiIiIiIqI+wk4eERERERFRH2Enj4iIiIiIqI+wk0dERERERNRH2MkjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1Ef6vpNXqVSQzWaRTCb1RdtOsVhEMpnE5OSkvoh2gHQ6jYGBAdRqNX2Ri23bmJ2dxeDgYKC8ZrlYH786wLZtDA4OYnh42BW+VRYWFjA8PIxIJIJIJIJsNgvbtvVoO9rs7CwqlQoAoFarYWJiAgMDA3q0bSfovtZqNUxOTvZdvhERbaRKpYLZ2Vk9uK1ufrPb9HUnr1qt4oMPPsD09LS+aNspFot47733sLS0pC+iPnPp0iW8++67sCxLX9SC5WJ9dkodUCwW8fzzz2NxcRH1eh35fB7T09OYn5/Xo+5YExMTuP/++zEyMgLbtjE/P4/p6Wk0Gg096rZSq9UC72ssFsNTTz2FZ599FtVqVV9MRESaYrGIGzdu4OjRo65weWMtEolgeHgYCwsLruUjIyO4//77kU6neWPNj9gFAIhEIqEHb6l6vS5mZmZcYeVyWQAQ+XzeFU79Z2ZmxjOvWS42xnasA1SGYYhcLqcH99xWlaFMJtNSroUQIpFIiO14GfJKpzD7urKyIgzDEPV6XV9ERERNhUJBpFIpPVikUimRSqVEPp8XqVRKABAARLlc1qOKmZmZbX1930p9/SRvO7t06RJ++OEHPZh2ibvvvlsPAlgudqVqtYpGo4FbbrlFX9RTlUoFH374oR684WZnZ3HlypWWu7TbVS/SaWhoCCdOnMATTzyhLyIioua17/Dhwzh37pwrvFgs4pFHHkGxWMTx48dRLBaRz+cBAG+//bYrLgDn2sJXWlqxk7cFisUixsfH9WDa5bZzuchmsygWi3qwi23bSCaTHKYW0k8//aQH9Vy1WsWvf/1rPXjD2baNkydP4oUXXtAXbUu9TKdHH30US0tLHc8bIqLdaHx8HKlUCrFYzBX+0EMPIZ1Ou8KeeuopAMDa2porXDpy5AjGx8c7zomw2/S8k2fbNrLZrGscrXzRXlWpVFyTDOh/3cYNo1qtIp1OO+tJp9OuAqJP2DA5Oekc18TEhLKm/yUn1ND3MRKJOJNnHD58GGgWcBmuk+kYiUQwMDDQlw2FTukvdUrTbuNKlUrFM16736vhlUql48QMk5OTzn4NDg7igw8+aFm+XcuFbdv47LPPcPjwYd/tyQ7ep59+qi/yNDs766oj0ul0Sz1RqVRayofegdTrBr/6xot+fkObFKdSqaBarSKZTDr5pm8fzXe21P3U//z2R5a7/fv3A0q+R5Q6Lejx1Wq1lnpXvr+wsLCA0dFRNBoNLC0ttWwDPueivh01vWR+RyIR3zKB5pPpRqOB+++/X1/UVqfrSDf5pKel/hcknaSFhQXnfJbpoYvFYjBNE6+88oq+iIhoV6tUKlhaWsLo6Ki+CNFoVA/CjRs3AACPPPKIvggAsG/fPqB5zSGFPn5zPer1uojH4877DCsrK8I0TQFArKysOPH0d4zK5bIwDEMAEIVCwYkXNq4fr/dx5L6VSiUhhBClUkkAEKZpinq9LlZWVkQul3N+m8vlRKFQEOVyWcTjcc+xwfl83hUu37sCICzLcuLpx6SH53I5kclkRKlUEqVSSRiG0Xfvd3RKfylomoaNq5Nx9e3LseD6mPF6vS7QLIP1el3k83mnXOoymYwwDMNzv9QysJ3LhTy31fTVlxmG4TrP/ZRKJWGappMnhUKhZb2FQsGVZisrKwKAMAzD+Z0My2Qyzr/90gRaHaCf39LMzIxznDMzMyKTyYhyuezkmV6P1Ot1YRiGSKVSol6vC8uynN/L/erEL9+DHp8Ml++81et1p96V55fwSANJnnvy9+oxyDpWT69UKiUymYyAx7mhkvvhx+s9tyDXkbD5pKdxuc11xOv3QtnXQqEg8vm8KJfLThroeSfJ33Sqf4iIdhNZd8prfCfyHb120GzD0f/yv/p2IZ/Pt1wcZQNCzRzZIFTJBrreMAoT14/XRTuRSLRcmGUjRp0gAFrDUng0GITS6NcLodowkLx+r4bH4/G2HZ3NhGZnJMxfkP0Mkv5h0jRMXD+y0aimvWVZnuu1LMuzTEFrsHp1YIRSwYXp5G11ufDqzHmFdZJKpVomGcnlcs5xyI6Tng6JRMK1Ha/jl2VITxN41AF+4XK9+j565a/cnnrsMr+CXmz88j3o8SUSCd9jUMu917HKtNbrUlnu9Q4ltHN8ZWXFtVwlz8l4PK4vcnilaT7gdSRMPoW5jnilk2hTl/jFF8o21GsKEdFu59Xe8rKysiJSqVSgNoasozvF2016OlzzwoUL+M1vfuMK+9nPfgYAmJubc8KWl5edR6vSAw88ADSHHanCxA2qWq1iaWkJiUTCFX7HHXcAAN59911X+L59+1rGDANwvZz/9ddfAwDuvfdeJQbw4IMPAgC+++47V3g7jz32mOfj6q+++koP2nDNGwGh/kZGRvTVuARN/zBpGiaun6effhoAcPXqVSdMDrmam5tzlbf5+XkcOXLE+befV155BaZptqTJgQMHXP8OYqvLRTQaxeLiIkzTxOjoKCqVCpLJJCzLwrVr1zA0NKT/xNPAwACmp6ddQ+p++9vfOv8/Pz+PRqOBRx991AkDgMXFRaytrTnbueOOO2AYBm677TYnTi8nL/nlL3+pBwHNYSbS559/DjQn2pBkXgf5REY7QY5Pnkt6WTx+/DiEEC3vNeiuXr2KRqOBhx9+2BUei8WQyWTQaDRaPuOQSCScYxwaGvIsk1DOSb/lfoJeR6Qg+dTL64hXmnYapszJlIiI/tfy8jLQ4fowOTmJvXv3Ym5uDo1GA6Ojo21fD5A24z33naJnnbxarYZGo4Fjx4653nPYu3evK460urrq/L9qz549elCouEHIxsfevXtd+3rs2DGgzfaC+Pbbb/UgwKNx1o1+aSiETf8waRomrk52KtR3aKrVKgYHBwGt83f58mXPxp6qVqtheXnZ+b3q9ttv14O6tpnlQu3o7d+/P3QHD8pMWHv37kU2m0WtVsPQ0JDTcZCdfK8bK6p0Oo21tTXEYjHnnbQLFy7o0TaFV0fBMAw9KJQgxye/n9hteXrvvfcAADfffLO+yMlT/abXRgp7HQlDr1ekbq8jqk7fz1vvbJ1ERLuNvFlZKpWQSCTQaDSQzWY934FWXb9+XQ/atXrWyfv+++8BAKVSqeXJjvyTjbZUKgXLslwX6x9//BHweKkyTNyg5FOder3eso9CCN/GQDsjIyMwDANXrlxxhcsG+EMPPeQK382Cpn+YNA0T1080GkUqlcLy8rJT3mZnZ/HWW2/BNE289NJLgDKBQyfynCC3oaEhLC8vI5FIYHp6GqZpdj31ca1WQzKZxNjYGA4cOIBMJqNH2VDyiZP6JEdegDrdBAhio4/v3//+tx7k8PvMx0YKcx0JYyOuI0REtPEOHjyIxcVFp6On3nCn9nrWybvpppsAAB9//LG+qMXJkydhGIbTI7dtG3/+858Rj8dbGkZh4gYln+p88skn+qJ1mZqagmVZzsyb1WoV09PTyOVyXTVMtgP1bnrQP31WPl2Y9A+TpmHi+nnyySeB5pBB2SCUnT/LslCpVPD222/vmG9+9ZqcVdGyLJTLZWfoptdshu3EYjEsLi6iXC4jHo9jfHy8ZcbaTuWoWq0iHo8jFothdXUV6XQ60BPbXnr00UcRj8dx6tQpJw1Onz4NwzDwpz/9SY8eSpjjW++dy3ZDfoOeO70Q5joSxkZcR8KQw8aJiKg7zz//PBBg9JIcik897OQNDQ3BMAzMzc15PkpVx9EODQ3h/Pnz2LNnD2699VbceeedGB4exuLious3YeMGJafzfvPNN/VFgLavYaTTaWQyGXz++eeIRCI4dOgQTpw4gbNnz+pRdwz9LnqQP/39M12Y9A+TpmHi+jl48CAMw8Drr7+OS5cu4fHHHweUd8Zee+01fPbZZ4GGJ951111Ac0id1zmx06gdvGvXrmFkZMT1jl7Yjh6aT2DlOuQHUWWD2Oujp7ZtO+Xj0KFDQLNzv1Wi0SguXryIwcFBjI6OIhKJYG1tDdeuXVt35yjI8cmL2euvv64vAppPotuRTyKvXbumL3KedHXz/iiU8h9mZESY60gYG3EdCcOvc05EtBvF43Eg5PB7+VpBp0/yyJuF1MNOHpqNbMuykEwmXRlXrVaddz/QvFBXq1UUi0UIIbC2toapqSnPFzDDxA1qaGjImUxDH99bLBYDTdLhJZvN4sCBA1hcXIRoDjs8fvy4Hq2FvLO8W4RJ/zBpGiZuO5lMBpZl4fPPP3dNMBGPxzE3N+dM0NJJNBp1Jpc5ffq0a5lsQLcTtlzUajXnSap8qiq/AVcsFp2wbPPD5vLfQRrOegdPdnL1yViCdPT0PI9Go3jmmWecf8t3I6enp137JvdBTqDhNbFJpzt8vVatVvH00087k8IIIVAsFgPdBOgkyPGNjIzANE1YloVsNutals1m23YuarUaHn30UadTpV9s33//fZim2fKkK2iZjEajiMfjnsfRTtDrSBjruY7o6RLGF198AQRolBAR7SZjY2NAyNdavvrqKyQSCd/r66effgrDMHyX70r6dJvrIadTR3Oa7UQi4Tm9ummazhTp8k9+g07/nlCYuF7kt6agfatIflsKzWnCE4mEME3TNU29/K3+CQWvb6rJKcfl953kX6lUEuVy2XNaftM0Rbn5rSV1vYlEwhXf71ttO1mQ9O8mTYPE7UTmuz5Vuswfr3JXV75Npv5OPc5UKiXK5bIoFArOVL/xeNz1fbL1lAson7FQ15dXvg2m7p/fsejkea2fxyo5FX+7OJKsF2Q8y7KEaZquqfDzzannZRp5TV8vw+Q30jKZjFP/5HI557jb1QFe4TJd1f1R81edDl9+CkNuL5/Pi5mZGVEulzumgyTXodYnIsTxyXyV60g0PzWhfxpA7n+pVBK5XM45ZvmZj3g87oTNzMwIQ/lOodC2o4a3I8uuV1r4nTNBryNh8inMdcQrndT1qt8eVMuQ1zHG43ERb/MJCSKi3UjWnV6fl5GfTMjn8841cab5bVS9vpbqzU/26J/U2e162skTzYSWjRY0G6D6xU9+68nvT73gh4mrUxuK8k82jESzkMmGlGwUyQLl91sZX/2TDR61YaL/6Q0Uuf5EIuHaD329eljC53tMO1G79JfCpGmYuJ14pbNlWZ4dba98alfO8s0PKZumKQrNj6lL6ykXslOnNsAzmYzrWGSDHgE7eKJ5Tsv9aUfur56HulQq5ToumSa6mZkZp2FtmqarcS2a25PL5f7JtJHp4Xcee4V7patfuExT+f02vz+9o6XT46vrDnJ8Urlcdsq/X3qWSiVhGIZnWpZKJdf5o9fbXunltQ2d7PzrF/J2aSp/1+46ov8WHfIpzHVETye/9XqliXruyfOx3TWKiGi3SiQSnm2qQqEgjObNcSg379u1LWTbJmi7ZreIiP9cMDdVpVLB9evX8cADD+DGjRuu4XlffPEFrly5grW1tdBxt9rk5CTuuece3Hzzzfjqq69cQ6suX74M0zQDDY2j/xUmTcPE7Ue1Wg2maaJcLjvDTCcnJ/Hhhx+63j3KZrOYnp6GZVnrfm9st7NtG6dPn8bjjz+OH3/8EV9++aWz7Ntvv0WxWAz9iYl+Mzs7i5deeinUu3m9thXXEa9zj4iI/qNarWLv3r09aYskk0k8+OCDXb2i09f0Xt9Gk3f721HvZAeNu9UKhULbx8T1et3zjkW35F1i/Q6yvOusLpd33PV/b3dh0jRM3H7l9SQvn8+7nibJp1jyKQl1r94cVtju7mIul+v4BHQ3yGQyLU/zNstWXEfqzeGd7coGEdFuVygU1t02K5fLPa/D+0VPJ14J4vz5821f3K9Wq87semHibjV90gPd/Pw8nnvuOT24K/KJjWVZEEIgk8ngzJkzqNVq2L9/PwDANE0IIVAoFDA+Po5IJOL6904QJk3DxO136nT64+Pjzgyh8inm8ePHMTU1hUQi4fmhdgpmfn4ey8vLerDDtm00Go1d/RRPmpqaciY+2WybfR2xbRvPPvss3nnnnUATuxAR7VbpdBqPPPJIxzacn2q1itdeew1vvfWWvogAbPpwzYmJCZw7dw6maeKZZ55xhtfduHEDH330EaBMGR4m7lYbHh52PvD84IMPOlObf/XVV3j33Xdx5MiRllnqujU5OYlvv/3W89grlQr279/vPP7u9O/tLEyahonbr2TnP5FIYGlpCQBQKBSQTqedfEdz1sZPP/0Uhw8fdn67ydVAX1hYWMCvfvUrGIaBTCaD//N//g9uv/12/Pjjj/j444/x+eef46233mJDX1EsFnH77bd3/MxKL23mdaRWq+Hq1at49NFHme9ERAFVKhXcuHEjVDutWq3in//85679bnEg+qO9zVBQZheUf6lUqmUygLBxt1K9Xhf5fN41eYGcTKTXw7UymYzvhA7qcM0g/97OwqRpmLj9ymu4Jm2scrnszAQmy11cmTGVtoedch0hIiLqlU1/kkfr1+6Ffv1JXad/U//wmniFiIiIiHafTX8nj9bv0UcfxdLSkuv9lm7HM1N/kB08ANi/fz8qlYoehYiIiIh2CT7J26HUd6zQfPfq9ttvbwlT37vS/82neURERERE/YedPCIiIiIioj7C4ZpERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iPs5BEREREREfURdvKIiIiIiIj6CDt5REREREREfYSdPCIiIiIioj7CTh4REREREVEfYSePiIiIiIioj7CTR0RERERE1EfYySMiIiIiIuoj7OQRERERERH1EXbyiIiIiIiI+gg7eURERERERH2EnTwiIiIiIqI+wk4eERERERFRH2Enj4iIiIiIqI+wk0dERERERNRH2MkjIiIiIiLqI+zkERERERER9RF28oiIiIiIiPoIO3lERERERER9hJ08IiIiIiKiPsJOHhERERERUR9hJ4+IiIiIiKiPsJNHRERERETUR9jJIyIiIiIi6iN93cmr1WqYnJzEwMAAKpWKvnhbsW0bs7OzGBwcxOTkpL7YRY273Y+rH1UqFWSzWSSTSX1R12zbxuDgIIaHh/VFgdRqNUxMTGBgYEBfROtULBaRTCZbzstKpYKBgQFMTEy4wreCbdvIZrMYGBhAJBLB4OAgFhYW9Gg7mm3bmJycRK1WY90OYHZ2tmMcIqKdoFKpYHZ2Vg9uq5vf7DZ928mzbRvz8/O4cOECGo2GvnjbuXTpEt59911YlqUvanHp0iVcvHgxUFzqrUqlgg8++ADT09P6oi1Tq9UwPz+P6enpHVHWd5KFhQW89957WFpa0hdtG7ZtI5lMYmhoCGtrayiXyzAMA7/73e/0qDuWbds4ffo0nnrqKdxyyy2s2wEcPXoUN27c2BY3GYiIulUsFnHjxg0cPXpUX+SwbRsDAwOuG2UjIyO4//77kU6nYdu2Kz79R9928qLRKI4fP459+/bpi7al48eP4ze/+Y0eDDQ7Fuod2+PHj+Oxxx5zxaHNMTIygrNnz+rB6xaNRrG6uorPPvtMX9RRLBbrqqx3eqpAwMGDB/Hcc8/pwUCzLKytrW1IeQhjfn4ey8vLePTRR4Hmfn322WdYW1vTo66LfMq02WQn9ujRo4hGo6zbFel0GpZl8Vwmoh2pWCzivffeQzqd1he5PPvss5439YaGhjA6OoonnnhCX0T93Mnbie6++249CABw5swZPYhoXWzbxoULF/Rg2oGuXbsGNG8UbKRLly7hhx9+0IM33LPPPovh4WEMDQ3pi3aMjazb//a3v+HChQscuklEO0q1WsXhw4dx7tw5fZHLwsJC29EN8gkgb3a1Yidvm8tms9t6qBjtPPLJiNddsY0gt1etVvVFLtVqFclkksMuQvr3v/+tB/VcsVjE+Pi4HrzhKpUK5ubm2g7j2al6VbdHo1Gk02kcOXJEX0REtG2Nj48jlUohFovpixy2bePPf/4z/vKXv+iLXI4cOYLx8XHUajV90a7W806ePgHA8PCw5x3GSqWC4eFhRCIRz79u47YjX2iPRCJIp9OuwqBPXDExMYFIJIJsNuvEqVarSKfTznb1dcBjX/2OH827DnJ/BgcH8cEHH7iWJ5NJ592v/fv3IxKJeN6pkI1juZ5OjemdaGFhwUlXOdmFV2dAzWP9T50oxbZtJ7/1eH7p7EdOABGJRDzfjwlyTlTaTOZSq9Vc5U7/09eFZnrJdFA7TrVaDffddx+Wl5cBoO06euXGjRv49NNPMTo66ls2q9UqRkdHPfNUV6vVnPI+MDCAdDrtOk+lYrHoKjPZbNa1fr0MtCtXXrwmZNHrEb98UAUtszqZBrKjIH8j9yfM8bVLq8nJSRw+fBhoXpS99kuvyyYnJ1vSWp1QpFgsYmBgAMPDw577I505cwaGYXR8irfb6/YDBw7Asqy+m2yHiPpTpVLB0tISRkdH9UUuzz77LP7yl7/g5ptv1he5yOH7ly5d0hftbqKH6vW6iMfjIp/PCyGEWFlZEaZpCgBiZWXFiVculwUAJ165XBaGYQgAolAoOPHCxvWSSCQEAJFIJEQqlRL5fN7ZJ9M0Rb1eF/V6XeTzeWe9+XxeZDIZYRiGMAxDCOVYSqWSEEKIUqnkWoeMA0BkMhnn33IdMo4k118ul4UQQszMzAgArmMVQoh8Pi8AOPH08JmZGZHJZES5XHbWkUgkXHF3ukKhIOLxuFOGcrmcACBSqZQrnp5WappaluWKm0gkhGmawrIsUa/XRSaTcfIzCJnOuVxOFAoFUS6XRTweb8mrIOfEysqKc0x63tXrdWEYhkilUqJerwvLspztyHImybJeKBREPp8X5XLZOS61TKlxN0u7c0Eui8fjLcu8qOlpWZZIJBIt6ZbJZEQ8HnfyXZaFeDzuxEmlUsIwDCcf/NJKr4NE8/xPpVKucL0eCZIPYcqsH7+8DHp8QdLKKw2keDzu+3uZnzMzM065zefzIpVKtZwHOsuyBDzOCYl1+/+S+aPXCURE25G8Hun1n0peZ0WHa5Ak6236X60tg3XI5/MtFyF5wVQb5PF43LnASvLCpl+kwsT1IhsCMzMzrnC1waHHlWGWZTkNl0Qi0VK4ZMNcrtvroi3jqGGFQqElTPg0wrzWqYbraeDX4Nto8gQM+9eJ7OToxy/zTzYQ6/W6gEfHT+34SLJMqmVCbgcBG9cAXA1o4VMJBT0nhNJxVMny47UdvTLzOlbhs96tKCdenTmvsHZkY1uNW6/XXccn00ctMzJ/1TTT00WWIT2tvPK1XbhMW9lpkPR1hymz7fjlpd/2ukkrv2OV9ZB+zrSry+T26/W6bwdPKOdJLpfTFwnBur0FmnUSEdF2J+tpv+t+vV533WjzuwapZB3Z7rqy2/R0uOaFCxdaZhH72c9+BgCYm5tzwpaXl1tmRnvggQeA5tAaVZi47egvvl+8eBEAcPnyZVc4mjOcoTlrYSwWQ7VaxdLSEhKJhCveHXfcAQB49913nX8bhoHbbrvNiXPLLbc4/y+98sorME0TIyMjrvADBw64/h3E448/rgcBzWE+m2lkZATNmwah/jqZn58HmutXmaYJAM5Qta+//hoAcO+997riPfjggwCA7777zgn78ssvAa1MRKNRp5x9//33Tng7+/bt8xxG9sUXXzj/H/Sc8PP5558DzRmkJJkWfi8ie81S9emnn+pBm25oaAjXrl2DZVlIJpOoVCoYHR2FaZpYXFwMNHHITTfdBAA4ffq0ExaNRl1p/Nprr7WcX9FoFGtra1hdXXXCTNPEL37xC1ecXjp48KAe5HoHK0yZ7UaQ4wuaVn5ef/11xOPxlncq/vSnPwHN8q87deoU0NyO1/kjyfPUqw5VsW7/X5v1ni0R0XrIV0a8rktoDtN89dVXfZe389NPP+lBu1bPOnm1Wg2NRgPHjh1z3lmIRCLYu3evK47k14DYs2ePHhQqblCycSELWjuyMbZ3717XsR07dgxQ9i+dTmNtbQ2xWAy1Wg3ZbLalkVOr1bC8vIzBwUFXOADcfvvtelDX+qWQV6tVNBoNV7pHIhGng/Ttt9+64uv/lrwaZDdu3NCDAKUj0S05EUbYc6Idr3iGYehBvrZL40/t6O3fvz9UBw/Nxnkmk8H09DQGBwdRLBYBZXYtNDvPXueXbnV11fn8gXwnbSuEKbNhBDm+oGnlpVqtwrIsz7yLxWIwDAONRsP3vbWNstvr9s1ObyKiXpLvFnvdKA3i+vXretCu1bNOnnz6USqVWp7WyD95tzeVSsGyLFfD9ccffwQAPPLII04YQsbdKPKOer1ebzkmIYSrE1prTgoxNjaGAwcOIJPJKGsK/pSI/qNWqyEej7ekufybmpoCmk+3DMPAlStXXL+XU74/9NBDTpi8a//RRx85YWhODmGaZtunC2GEOSf8yCdU6pM4OVGF1xO73WBqagozMzNYW1vD4cOHMTw83PbpRjuzs7MYGBjARx995DwB2ixhymy3NvL4OnU29BEY2xHrdiKi7cO2bTz//PP429/+pi+iLvSskyeffnz88cf6ohYnT56EYRjODG52c4rUeDze0nANE7cbcthfO/KO+ieffKIvcqlWq87QpdXVVaTT6XXfjd8pKpVKy9O2IH+d7NmzB8vLy21n4JOmpqZgWZYzw2W1WsX09DRyuZyrMzU0NOQ8DZJ3jIrFIpaXl/Hqq6868dYrzDnh59FHH0U8HsepU6ecjszp06dhGIYzJG4nkbNomqaJcrnsDN0Mkr+qo0eP4ptvvkE+n8fy8jJGR0ddN4KCTE2fzWZx8uRJXLt2DVNTUz3r3IcRtMx2I+jxBUmrdjoNBVaHOG6m3Vq360NFiYh2ikuXLsGyLNx6662utuL+/fsBZYZnr9mIJflKF/Wwkzc0NATDMDA3N+fZYJPDqtCMe/78eezZswe33nor7rzzTgwPD2NxcdH1m7Bxw5D7ODY2pi9qcf/99wMA3nzzTX0RoBzboUOHgGbDzc9dd90FNBtWXum0U41s0Dt5cnpd+W6eTi1X6XQamUwGn3/+OSKRCA4dOoQTJ044Q9ZUL774IlKpFJ5//nlEIhG88sorKJVKXQ8P8BLmnPATjUZx8eJFDA4OYnR0FJFIBGtra7h27dq6OwGbTe3gLS4uYmRkxPWOnlcatRONRnH8+HHMzMyg0Wg4ZUQ+qfVK30qlgmq1imKxiOnpaZw/f96387MZwpTZMIIeX5C08iOfRDYaDc948sl4t+X0nnvuAdoMZ/Wzm+v2MEO4iYi2SjweB/7/7d3fjxvV3T/w9/4BQGf3rkWoWi8XFVQbFW9SoQUpkchsU1RR8cMOqh7lAjWxQZUeSnbBS/SogoR4m/YiotldlEgVEthuVkqE8JJdpEYPHhAEg9ZqIy5gLNQn6pUdN/QPON+Lzme+Z45n7LHXu9l13i/JSvbMeDxzzpmZc2bOj5CuKA8//DDy+XzbR1pO2LaNfD7fsSK32S43w2RglTx4BRYpsOkJV6vVcPnyZf/vYrHoF7SUUrh58ybOnTsX2rejl3V7ceHChci3IeYNenJyEolEAqVSqW2urWKx6Df5CRsIQ5peibGxMb9gpQ8eAa0ZaifmCRFHsVgMfXsm4fV63V9WLBbb/r6dpMnasWPHsLy8HFiWzWYDfV2y2SweffRRXLlyxW9qJQMt6Jre5NynTp3CN998A6UUPv/884FW8ETccyJKrVbDc889hytXruDmzZtQSqFYLHYsuPejn3zVC7OCJ+ev3kcvTkXPcZy2fGBOlC2TQmez2UDlo1ar4b//+78xOTkZOqhJt9/eCnHzbK/iHl+cuArT9FpVvPzyy4DXLFQn/dNef/31QDhiXucAYN++fUAfefNOurYLSbs4FVsiottNrlVmM/fp6WnMzs62fWQgqgMHDmB2dja0xcK1a9dizat6RzGH29yMhjcnGLzh8W3b9qdA0Ic0TSQSyrIslc/n/Y/MNWYOxd3LumFkLqukNs+azDmkD1Huuq4/x1LYEK0y1LsMU21786zpQ7/L8K0yt5HMPwVvGHDZrr6tVCqlKpWKKhQK/veTyaQ/dLcMyZ1KpVS5XPa3IcelDy8uQ85CG/q7UCgEhkJPJBIq783dJekky2Xobv1vc5j+20HiQPbNtm1lWVZgiHGZUyvpzaEmn3K5rCqVSmCYXj1O9XUrlUrbcOZhZBh/y7ICeVDylT7sb9xzQrYJYzh6GXpd8k8+n1dLS0uqUqkEvq+nvT50v75dfX19SPdCoRB7uP5+xJkmQfJjp3WUtt7S0pK/Xt6bB02PNzmXouJctmNZliqXy6pQKPjnlIRJPIbN+xYV3ks69JJno+jXEn0qgV6Or1tcKW1fE4mEf21TRv6W33e9uRz1qSEajYb/+ylvzsc4Ut5cf2Hu9Gu7TqabMKftICLaieSeGHY9CyP3tLBruNKmCIqacudONdBKnvIiWgqQcgPTCwtKm18o6qPfoHtZN0rBm0w7ap8k8+gfO2TS2Y2NDf9mLZUMvbCyoU10bdu22tjY8Ldt23agEGpuSyoZiURCFQqFwHb1AoZkZPMTdgx5b462qDiSQkbcv2+ncrnsx61UVE16GpsfvdDquq5fEAv7mIV5Xd6rCOsfiWczXHQ7J8K2WfEqm1J4i/pIodMMt725v8xw2a4UfCXvbSXJ61FxKgqFQtvcYKaKN+m85AU5VvMa02g0VC6X89M5bB25tuhxYHsPEOTGE5aulUolMtwMi0oH+b24eTZM2HahpXGc41Mx40oZ89zpyyV/y/fDzk9zHxHzuiJxGrY/6g6/tusymYxK7IAHckREcdm2HXgY2IlcB81rn5DyapyXP3eSERWnY9SAOY6DTz75BA8//DBu3LgRaFr05ZdfYn19HTdv3ux5XWo3MTGB119/PXSQmmKxiMOHD/t947r9vdMtLCzgwQcfxN13343r168HmlNdvHgRiUTCb36azWb91//6cLu3bt1CqVTC8ePH25oB3g7NZhMnTpzAs88+i++++86fOwxeX6VisYirV6+yecIu1UuevVOl02mMjo527A93J2s2m7j//vvx3nvvhTZhIiLaiWq1Gvbs2QPXdfvuuy1mZmb8ppykMWt9W02ecnYiy3tZl8LJm4Qw5pu6bn/vZIVCoeNr+obXXEx5T707NWsql8uxmxBspYbXFK7TG7BcLhf5lqNXea95rqQ7tDfl8sZB3izof1N/esmzdzI5DwaVz4dN3mvmS0S02xS8rgSbIWUSajfQgVfieOONN9o6v+tqtRoOHDgA9LguhTty5Ajm5uYCnfqz2WxgnWHQ7ZhWVlbwwgsvoOYNUd/J+++/j6eeesoM3nYrKysdJ3RuNptotVoDeYtXLBYxNzcH13Xx0UcfQSmFfD6Pw4cPo16v48qVK/7oVtPT0zhy5Ahs2941b3l3orh59k43NjaGK1euYG5uLnQUzztZsVjErVu3+PSaiHaldDqNJ554ouv9MEqtVsObb76Jd955x1xEALa9ueb8/DxOnTqFRCKBX//6135TpRs3bviTU0uznF7WpWgLCwuYm5vz/3ZdF9euXcPhw4f9sEKh0PHvbc4mPZuamkK1WoVt2zhw4IA/vO7169dx6dIlHDlyBOl02p9cvdVqIZPJYHJyEg888ADgNdu8ePEizp8/P5CK02atrq7i5z//OSzLQiaTwY9//GPce++9+O677/Dxxx/jiy++wDvvvLPpkWbFwsIC3nrrrcAE0CMjIygUCn5z35mZGaytrcG27U1PY3Kni5tn6T+azSZWVlbw2GOPbbppzzCQOT63YlRgIqLt5DgObty40dM9r1ar4dNPP90RXWt2LPPV3nYoaKONyUdGGDP1si7duRqNhsrn84FBGGQABbOZl+u6/kAFsm4ikVC5XK5j08jboVKp+CMMyr7qI/QNkjTX1EnzTR0iBq+g3vSSZ4mIiIh6se1v8ohoZwp7kzcxMYE///nP/oAOExMTWF9fRyKRQCaT4Zt0IiIioh1o2/vkEdHO5bqu339TRnWUCt7MzAzW19cxPj4O13WxuLiIhYWFwPeJiIiI6PZjJY+IfIlEAgcPHsTIyAgOHz7sv9XLZrNYW1vzO0fLv3Nzc313mCYiIiKircHmmkQERDTXJCIiIqLdh2/yiIiIiIiIhggreUTkT7Phui4mJibMxURERES0i7C5JhERERER0RDhmzwiIiIiIqIhwkoeERERERHREGElj4iIiIiIaIiwkkdERERERDREWMkjIiIiIiIaIqzkERERERERDRFW8oiIiIiIiIYIK3lERERERERDhJU8IiIiIiKiIcJKHhERERER0RBhJY+IiIiIiGiIsJJHREREREQ0RFjJIyIiIiIiGiKs5BEREREREQ0RVvKIiIiIiIiGCCt5REREREREQ4SVPCIiIiIioiHCSh4REREREdEQYSWPiIiIiIhoiLCSR0RERERENERYySMiIiIiIhoirOQRERERERENEVbyiIiIiIiIhggreUREREREREOElTwiIiIiIqIhwkoeERERERHREGElj4iIiIiIaIiwkkdERERERDREWMkjIiIiIiIaIqzkERERERERDRFW8raY4zhYXl42g3eN5eVlOI5jBt+x6vU65ufnMTo6GogXx3EwOjqK+fn5wPp3slqthnQ6jZGREYyMjGBmZga1Ws1cLVKxWMTMzAwWFhbMRUPLcRxks1nMzMyYi3qy2bineJrNJrLZLEZHRzEyMoKJiQmsrq6aq0UaVHp30mw2USwWMTExsavPJYmrkZGRQPjy8jJGRkZ6ivdBqtVqoftFRHS7sZK3hYrFIm7cuIGjR4+ai3aNo0eP4saNG1tSeZFCvBRER0dHkU6n4TgOarXajiuQ1Ot1rKysoFQqodVqmYtJU6vVsH//frzyyitQSqFQKODatWt44403zFVDFYtFXL58GWtra+aioVWr1fDBBx9gcXHRXNSTzcY9xdNsNjEzM4PJyUncvHkTlUoFlmXhV7/6lblqKMdxBpLe3aysrOAPf/gDXNc1F+0ajuPgk08+2fK46lWtVsNf/vKXHbdfREQAAEVbolAoqFQqZQbvWqlUSuXzeTO4LxsbGyqRSCjLslQ+n1eu6yqllGo0GqpQKKhEIqEADOz3Bi2VSikAqlKpmIu2RaPRUEtLS2bwjpJKpVQymTSDI4UdU6VS2bZ8UKlUblt6mgAo27bN4Ng6xf12xOWdYmlpSQFQjUbDXBQpLP43m95x5PP5bTuXtpJlWWonFlsA7Mj9IqI7G9/kbYFarYbDhw/j1KlT5qJd609/+hNOnz696aab8pbh5s2buHr1KmZnZzE+Pg4AGBsbQzqdxmeffYZkMom//vWv5td3hJ/85Cdm0La6cOECbt26ZQbvKOvr6xgbGzODI93uY3r99dfNoF0rKu6bzSZOnz5tBlOfrl69CnjXrTgcx9mx17TdYu/evWYQERFFYCVvC8zNzSGVSvmVl2EgFbAjR46Yi3ry3HPPodVq4dy5c5icnDQXA95vlUolM5i8Zoxzc3Nm8I7TS3PW231M2Wx2qJqFhsW9NC0MW0b9+de//mUGRarVavjFL35hBhMREW0ZVvIGzHEcrK2tYf/+/YHwhYUFv++ZfBzHaQsfFHO78ntC7wsXt+/bo48+Ctd12zq4T01NYXR0FPV6PRBuWl1dRbVahWVZSKfT5uKA8fFx/PKXvwyEmYMcTE1NoVgsBtYRcdbVB1EBgPn5eYyMjCCbzfrrmNuZmZnBl19+qW3lP8IGCTG3v7q6iomJCX87zWZT28J/1tcHy5iYmAgM2rOwsIDDhw8D3oME2Y4w93VqaqrtzWu9XvfTXvpA6sfbiTmYh/Sf1MkyAFhbW/P/NtcT3Y5JyLHJfptpiYj965YnZ2Zm/P40jzzyCEaM8yEsTs3fbjabWF5exsTEBBzHaUvnQQ140u34ouK+WCxi3759qFargfWi0kQ0m00sLCz4xxI1cEecODLPBf1cM+OvVqv5eXRiYiI0/lZXVzE1NYURLz/Mz88Hzid9QBOp4EpcdBLnWGRb8mBA4jMsbuDt6/79+9FqtQLpEmZhYcH/7bB+0GH71y0dTY7j+HHX6fthaS9x7DiOfxwjWl4yw4WZ/uY5Yl4LEZLfJa/ozLwj9PRHjHiFN4CL7JP5CbsmhWk2m/5xhuVLYebfbDYbOJf1+7Mel3p43H0iojuY2X5zt3Nd1+8zBUClUim/z9d2yGQykf21KpWK36dA7xth27ZKpVI99e2IQ/qMAFAbGxvmYpVMJtv6QXUifaQymUwgvNNv6CRd+ul/4rqusixLZTIZ1Wg0VKPR8OPa3J846zYaDZXP5wPpkclklGVZyrIsf51kMqmSyaRyXTewHT2Ny+Wyf2ySrub2C4WCyufzqlKp+NvQ80Cj0VCJREIlk0l/n5PJZFteiuqnJutLuPR7NNNFX8d1XWXbdqz0KJfLCoCfX1zX9fevUCiYq/eUzlHHJOG5XE5lMhlVLpdVuVz200g/X+R4y+WyUtr+JhKJrueV9Fcyz9k4+ch13UA653I5Zdu2yufzfvyY+9pNWNz1cnxh31fedQYxL/mSn+TYlXb+5nI5f704cWSeC+a5trS05MfV0tKSymQyqlKp+Ncv81gKhYJKJpN+vs7lcgretV55cSVhcm2VferUTzrOseh6iU/VIV0kPJfLqUKhoCqVSui5H/ccDyN53LZtfxuy/2Hf1697SruXyPVJefsj2zD3U/ZL/u7lWqi8Y9PTQr/e6HGu5x3ZBzP9u8WrCrkG6PfOOOUHWVfyWi6X84/XzHOyTH6r4pULLMsKpIO+D/r+hsUXEVGY+HeoXUBu0nJhlI9lWbEu1IMgNxGz4CXMwtnGxkbgxjlochM2b+KNRkNZXmWmFxKfuo2NjbabZpjNDKhi27ZKJBJmsL9N/fd7XVffJ9d1/bwiBVEzbcK2U4moqMj2pXAuYBT4wr4veUUPC1tPeYUUswAp39cLv2belIJaJ5JXzIKu67p+fjDjyDy+TqKOScLN88MskCkvns3vS0Gv24OMsO2pHvNRVGVArgfmvnUSFne9HF/Y91WPlZK8VyjX413SQz/GXuKo07kmaaBXIPXvCMmLZlpJPOvXORgF5I2NjbZ8quvlWFTIvnUTlS7wziF938POiTjneBSJX3M9vTJkrmveM8MqZJ3OHTNuJKzbtVB5cW6GyfXG3G7UPki86scRFq+NRiM0bmR/wx5imWS/9DTUyyOyb/L75jYLhUJoPJgPMOQhCBFRHEPVXPPMmTOhfU5arRbOnDljBm8JaRIV1Rn/0KFDyOVycF0XJ06cwJNPPolSqdS2vt5UajNeffVVwGuKoltZWUEmkwmEFYvFjs13hBnHk5OTmJ6eDoSF6XcI73q9jrW1NaRSKXORP2DGm2++CfS4rm52dhbwmomOj4+j2WxicXER6XS6LW0OHjwY+DuOQ4cOmUGBfmB33XUXLMvCgw8+6Ifdfffd/v+7OX36dFvz1h/84AcA4PdvvOuuuwAAJ06c8NcZGxtr+57pww8/RKvVwuOPPx4IHx8fRyaTQavVwsrKSmDZID399NNtaQAA169fB7xmXWtra7BtO7D8vvvuAwBcunQpEB5Hr/nohz/8IQDg2Wef9cMA4He/+x0A4OLFi4HwXmzF8XVz+vTptrw/PT0NpRTOnTsH9BFHwjzXdD/72c8Cfwu5Jkk+M683iUQCMM4pALBt2193cnIyNB9hE8cyKHv37g3to6w3DY9zjndjDhp18uRJWJaFtbU1v1nhW2+9hWQy2ZY2L730EuDtx2Z0uxaurq7Cdd22/t/j4+N+Ose1d+/etuMAEBgA56uvvgJC4ubAgQMAgH/84x+B8E70NBwfH/enLfnggw8AAO+++y4A4LHHHvPXA4B0Oo1EIoG1tbVA8+STJ08imUyiVCphfn4e58+f988/IqJuhqqSZ/ad0HVatt1efPFFJBIJLC4u4vjx46E3oQsXLuCtt97qu2IkpqenYds2FhcXA23+f//73+OZZ57x/3YcB6+++qpfSe2mW0VwkD788EMAwD333GMuwo9+9CNAK+T0sm4nsh0pvOvCwjZL5to6dOiQX8E3CzlR6vU6Wq0Wjh07FujDsWfPnsA6UilbXFzExMSEf050m8fx8uXLQESlUwo1W1HR6EZG45RC2p49ewLHf+zYMQDAN998E/heHIPKR1KgjXtehdmK4+vEcRy0Wq2u+XxQcdSLWq2GVqvV1l9Jfufbb781vxLL7TiWOGRwl7jneD9kxMqvvvoKtVoNruuGVobHx8dhWRZardaWXv/ff/99AMC9995rLtr0Q89OovJOWJ6I64EHHgAAfPHFF4BWDgmL32QyCYQ8qDh//jwA4NSpU/7/iYjiGKpK3m4xNjaG48ePA15lK6xj9uzsLP785z+bwX2RyoK8zXQcB5ZlBZ46Tk9PD7ywaJI3EVE30yidhtY3n373sm4nvTy9HRTptH///fcDPQzr/89//hMAUC6X4TXBbvvIg4Rz585haWkJN2/exOHDhzE1NRU6sIWu0yiCUoi5nSStGo1G23ErpfrK14PKR6awAZFGIgbhEFtxfIOwVXHUSb1eRzKZbIsD+fT7luN2HEsvejnHN+Pf//63GRSwHVMY9FtZ7df09DQsy8L6+nogXPKE+dZtM8xWMDrzTaKYnJz03zDLm0EiojiGqpLXacTGTsu2W71ex6VLl5DJZOC6Lv74xz+aqwyUNAWRt3nvvvsufvvb35qr9cRsLhWHNDUyb6ZxhY1qKcxmPL2suxPICIDr6+v4+uuvMTs7G/okO4w0w/z444/NRaGOHj2Kr7/+Gvl8HtVqFfv3749VsJLmkWEGUcDslzxp/+yzz8xFmzaofCTr3nfffbBtu+3TyVYeXydx53QbVBzF8b3vfQ/VajX0wdggbOex9KLXc7wf3//+9/3/X7t2LbDMpK87DM6dOwfXdf2RN2u1GhYXF5HL5QZybTO30enBmjTDFqurqxgdHfWbbZqjWxMRRRmqSt5LL70Ey7LMYFiW5fcn2GrS5KJToTmbzSKfz+O1115DMpnEqVOntrT5C7S3QmfOnMHi4uKmnk6GxXEcTz31FCzLguu6sZrPypDkUggOqxxKYU/6yfWybifSN24zfal6ceLECVSrVZw/fz60KU8nk5OTsCwLpVIptPAbFtdjY2OYnZ3F0tJS1z51UjmXyZ913333HeBNr3G7/PSnPwUAvP322+YiIOL4uxlUPjLXTafTuHLlStunk604vk7kAc7a2lrodUz69w4qjnohU9NE5dd+4+J2HEsv+jnH42o2m0gkEhgfH/ffarVardCKiL7uVnnooYcArR/bdkin08hkMvjiiy8wMjKCJ598Ei+//DJOnjxprtoT8/oob+Q+/fTTwHrQ3hzqb0vr9TrOnj2L1157DefPn4dlWfjVr34VmgeIiExDVckbHx9HtVoNdJ5PpVKoVqtbelPSSUFAmteYstksjhw54g8CIG3sjxw5sqUX7nQ6DcuysLi4iEwm03NFAtrTR7OwU6vVYlVSx8bG/OZU2Ww2tBAhisWiX9GanJyEbdtotVpthRnpSyOV+F7W1Zlxv2/fPliWhWq12radTk27+hVWmJYCQifNZhPNZhPpdBqu62JmZiawrVqt5vepcxynbQCebv3xoFXOS6VS236+//77SCQSA31TLscU1+TkJBKJBEqlUttcWsViseemt/V6ve98ZKaZrGsOyNKLQR+fMNNSJ4MypVKptt+T87bfOOolbU3ycOrYsWNteTmbzba9/Y77W/0ey2Z0iv8wcc7xXjmOg2q1irNnz/phL7/8MhAyWFe9Xke1Wg1tRq7n+2az6Tch7vUYoQ2+s7i42HaPiJuevcpms3j00Udx5coVvwm0DBC0GWfPnkUymfSvjy+88ALgddMwlUolZDIZv6zS9OZEzOfzGBsbw+TkJN544w20Wi08//zz5tfhOE5f8U1EQ8wcbpM2R4apN4c1V97wx+aw5Eob/tscKl6Gx9bDXNdViURCWZbVNhR1N1FDTQsZSjpqu7I/5vKw4aM7keGiLctS+Xw+8L2NjQ2VSqXahpiWeZP04dMr3vxCZlzHXVfiEhFD3Mt+yvJKpaLy2vxnqVTKj4uoeaRk+3qcSR7R40zSJplM+r8jQ3jbtq2Wlpb8IefhTcFR8eaZkt/S55Gyvfmw9KHZZfjupaUlfx/z3lD55nDpJomLpDF3Vthw9nq8xckTUcekz5OmnwMyX5s+5LmkObx8JcPhm+dUGNlfSU/JC3HzkQpJP31dc1qATvS8oadJ3OPrFPf6EPiFQqHtHNPJVAXdfi9uHHU718Lm4NPPH31b+jHKUPuWMcWH5HV0uN6Z4h6Lvq65b53o14JcLqdc1/XT2zLOQcn7CW0exDjneBTZnqXdN+TYzCH59d+RY3O9eerMaQZk/5PJpCqXy/4ceHLtknOql2uh0vKqZVn+HHeZTMbfRi6X89eXvKMfRy/xKtefpDd/oHzK5bKqVCpdrx9KKT8vZDIZ5WpzqoaljT7dSkObj9FM60wm0zalQsO7R5vHK+dEImQKECK6c7GStwVsbwJenRQCYcyFIzdD/aOMQgq0gorMvZP0JqvtxYY3eW4U/ffCCmJyIzIlk0mVSCS6VhR0ruuqXC4XKLQkEgn/Jhlmw6sAyvpSsAjTbV0zfhEyR5HyKrZSsEh4k1FLIUa2F5aGUdvX84F88vm8amgTC8vvSEE74VV+hGzDtu1AAUIKBrLdVCoVWF7xJgOW4wnbRiflcjmQXub2VURcIMazJPOYwrYTFadC/54UXuMU0JS23+Z3uuUjIfuvF0QTiUToeRQlKm+IbscXFmfQ4l4qWpb3cKUb13UDx27+nugWR93SzVyGGGltnpf68XSLx066HYuK2L7scyflctk/n+U6Ym5Dri1muOh2jndSqVS6HpvQKykIiWOdPOyxtAcatm2rjDepfVRaRh2/yGsTqOvXhUwm4x+z+f1O2w2LV0kz/bpmfsIqaqaGN+G75Mmw81PXbV19X/U4MfdNzokNbfJ4IiIxov5z4aABqtVq2LNnD1zX3bJmok1voI7PP//cXBRpeXkZt27d6qsZSrPZxP3334/33nuvr0FXiIbZwsIC5ubmUKlUeH4Q7TILCwt48MEHcffdd+P69euBJvkXL15EIpFoa8ZLRLTTsZK3RYrFIi5fvrxlN4ZsNovHH388dGLZKBMTE1hfX++r4imDoPRTQSQadqzkEe1OxWIRf/vb3yIHWWk2m3j++ee37F5ORLRVhmrglZ0knU7jiSeeQDabNRdtSr1ex/LyMp599tlYFTzpiO04Dg4ePNhXBa9YLPb9BpDoTrAVg/EQ0dbrdo9eWVnxB0whItpN+CZvizmOgxs3bgx09MG40uk0SqUSkskkWq1WX2/xZE6eOBVKojtRs9nEvn374LoulpaWYo1YSkQ7w9TUFKrVKmzbxoEDB/Dwww8D3rygly5dwpEjR27L/ZuIaLNYyRti0oTMtm3k83lMTk6aqxDRJjiOg0ceeSQQZtt217nviGhnaDabuHDhAi5evIhqtQp4c8Gm02kcPXqU900i2rVYySMiIiIiIhoi7JNHREREREQ0RFjJIyIiIiIiGiKs5BEREREREQ0RVvKIiIiIiIiGCCt5REREREREQ4SVPCIiIiIioiHCSh4REREREdEQYSWPiIiIiIhoiLCSR0RERERENERYySMiIiIiIhoirOQRERERERENEVbyiIiIiIiIhggreUREREREREOElTwiIiIiIqIhwkoeERERERHREGElj4iIiIiIaIiwkkdERERERDREWMkjIiIiIiIaIqzkERERERERDRFW8oiIiIiIiIYIK3lERERERERDhJU8IiIiIiKiIcJKHhERERER0RBhJY+IiIiIiGiIsJJHREREREQ0RFjJIyIiIiIiGiKs5BEREREREQ0RVvKIiIiIiIiGCCt5REREREREQ4SVPCIiIiIioiHCSh4REREREdEQYSVvizmOg+XlZTN411heXobjOGbwHater2N+fh6jo6OBeHEcB6Ojo5ifnw+sfyer1WpIp9MYGRnByMgIZmZmUKvVzNUiFYtFzMzMYGFhwVx0x0qn0xgdHUW9XjcXddVsNrG8vIyJiQme07fRwsICJiYmMDIygtHR0Z7yt+M4yGazmJmZMRcN1OrqKmZmZrb8d7aSxNXIyEggfHl5GSMjI1hdXQ2Eb5darRa6X0REg8ZK3hYqFou4ceMGjh49ai7aNY4ePYobN25sSeVFCvFSCRgdHUU6nYbjOKjVaj0VfrZDvV7HysoKSqUSWq2WuZg0tVoN+/fvxyuvvAKlFAqFAq5du4Y33njDXDVUsVjE5cuXsba2Zi6iPl24cAHnz5+H67rmItom2WwWX375Jb755hu4rouDBw9ibm4u1sMPx3HwwQcfYHFx0Vw0UKurq3j77bd39bnnOA4++eSTLY+rXtVqNfzlL3/ZcftFRENK0ZYoFAoqlUqZwbtWKpVS+XzeDO7LxsaGSiQSyrIslc/nleu6SimlGo2GKhQKKpFIKAAD+71BS6VSCoCqVCrmom3RaDTU0tKSGbyjpFIplUwmzeBIYcdUqVS2LR9UKpXblp7bKZ/P95R375R42Q4bGxsKgCqXy+aiSGF5H4CybdsMHig597b6d7aaZVlqJxZzAOzI/SKi4cI3eVugVqvh8OHDOHXqlLlo1/rTn/6E06dPb7qZl7zhuXnzJq5evYrZ2VmMj48DAMbGxpBOp/HZZ58hmUzir3/9q/n1HeEnP/mJGbStLly4gFu3bpnBO8r6+jrGxsbM4Ei3+5hef/11M4gYLwP16aefAgDuvvtuc1Eox3F27DVwt9i7d68ZRER0x2AlbwvMzc0hlUr5lZdhIBWwI0eOmIt68txzz6HVauHcuXOYnJw0FwPeb5VKJTOYvGaMc3NzZvCO00tz1tt9TNlsdlc3TdsqjJfB6uUhRq1Wwy9+8QszmIiIKDZW8gbMcRysra1h//79gfCFhQW/75l8HMdpCx8Uc7vye0LvCxe379ujjz4K13XbOqxPTU3FGgxidXUV1WoVlmUhnU6biwPGx8fxy1/+MhDWbDaRzWYxOjqKkZERTE1NoVgsBtYRcdbVB1EBgPn5eYyMjCCbzfrrmNuZmZnBl19+qW3lP8IGCTG3v7q66g+4MDMzg2azqW3hP+vrA5VMTEwEBu1ZWFjA4cOHAe9BgmxHmPs6NTXV9ua1Xq/7aS99IPXj7cQcSEX6T+pkGQCsra35f5vriW7HJOTYZL/NtETE/nXLkzMzM37/mEceeQQjxvkQFqfmb5sDmpjpHKe/FUK2UywWMTo6iqmpKTSbzbb8ZH5Xlsnx65+wc7xWq/l5YWJiIrCf3eIlTLPZDAwqMjExEfqdOHFqHqt+bprx1Ok4xOrqKqampjDi5Z/5+fnA+acPaNJsNv3tmftlinMsci2WBxkSn2H5HN6+7t+/H61WK3AOhVlYWPB/O6rftJkmvQ4Epl+XOn3f/J2FhQU/jh3H8Y9jRLsemOGinzQ2z3/JKzpzu0JPf8SMV9mOvv/yiUpbk37ehuVLYebfbDYbuLZJ3JhxqYfH3SciGiJm+83dznVdv88UAJVKpfw+X9shk8lE9nmpVCp+HwG9r4Vt2yqVSqlGoxFYf7OWlpb8eNjY2DAXq2Qy2dYPqhPpp5HJZALhnX5DJ+nSTz8P13WVZVkqk8moRqOhGo2GH9fm/sRZt9FoqHw+H0iPTCajLMtSlmX56ySTSZVMJpXruoHt6GlcLpf9Y5N0NbdfKBRUPp9XlUrF34aeBxqNhkokEiqZTPr7nEwm2/JSVD81WV/Cpd+jmS76Oq7rKtu2Y6VHuVxWAPz84rquv3+FQsFcvad0jjomCc/lciqTyahyuazK5bKfRvr5Iscr/Z1kfxOJRNfzKqqfWpx85LpuIJ1zuZyybVvl83k/fsx9jbK0tOR/J5/Pq1Qq5afh//7v/wZ+x2TbtkokEm35NJFIBNaTY11aWlKZTEZVKhX/OmGmV1S8hJH8J3GltPM9l8v568WJU/PcyRvnph5PcY6jUCioZDLpnwe5XE7BuzcoL+9ImFyLZZ869auOcyy6XuJTdTiHJDyXy6lCoaAqlUrotUJ59yO5tzQaDWXbth9vnci5J9c/OQfhXfvM7+vXSaXde+R6prx0ld/X97PhXfug5ete03hjYyOQFvr1KWq7sg9m+seJVzMt9XttnPKGrCt5LZfL+fndzHOyTH6r4pUjLMsKXNv1fdD3FyHXViK6M7SXFnYxuenKhU4+lmXFuvAOgtwUogp1ZuFzY2MjcCMcNLmpmhWwRqOhLK8y0wuJT93GxkbbTTDMZgZUkUKsSbap/36v6+r75Lqun1ekYGmmTdh2KhEVFdm+OdgCjIJK2Pclr+hhYespr9BhFnzk+3ph1sybUvDqRPKKWXB1XdfPD2YcmcfXSdQxSbh5fpgFLOXFs/l9KbiZBVJT2PZUj/koqnAv1wNz36LIvkjcyTVCSH7SmRVwpaUZjEKnbN/cz7DtRsVLmLxXKdPTSdJP/61e4rTTuSn7plcg9e8IiQfzGCRd9LiFUUDe2Nhoy9e6Xo5F9RifqsM5BO+c0/c97ByqeJUB83yX73ci25P7lJC8pm9Xjsu8x8o5oe9TVByY6aZ6SGPlxbkZV3J9MteN2gc5Lv04wuJV4tCsjMl+hT30Msl+6Wmol19k3+T3zW0WCgWFkPxhPsCQCjIR3ZmGqrnmmTNnQvsCtVotnDlzxgzeEtVqFfD6lYU5dOgQcrkcXNfFiRMn8OSTT6JUKrWtL82TOjUXiePVV18FvKYlupWVFWQyGf9vaRomzTrCmowIM44nJycxPT0dCAvT79Dt9Xoda2trSKVS5iJ/YIg333wT6HFd3ezsLOA1Ex0fH0ez2cTi4iLS6XRb2hw8eDDwdxyHDh0ygwL9ne666y5YloUHH3zQD4s7QAMAnD59uq156w9+8AMA8Ps33nXXXQCAEydO+OuMjY21fc/04YcfotVq4fHHHw+Ej4+PI5PJoNVqYWVlJbBskJ5++um2NACA69evA14zrbW1Ndi2HVh+3333AQAuXboUCI+j13z0wx/+EADw7LPP+mEA8Lvf/Q4AcPHixUB4N3Lejo2NRfZdFX//+98BAA888IAfNjY25g868c9//tMPF+Z+irBmcHGcPn267VyZnp6GUgrnzp0D+ohTYZ6bup/97GeBv4U0xZN8aV6fEokEYJyDAGDbtr/u5ORkaL7DJo5lUPbu3RuaL/Sm5G+++SYOHjwYOIaxsTFYloVWqxXZhFo3MTER+P6hQ4eQSqXQarXw2WefAQDeeustJJPJtrR56aWXAC9vbEa3NF5dXYXrum39xcfHx/10jmvv3r1txwEgMADOV199BYQMwHXgwAEAwD/+8Y9AeCd6Go6Pj/tTzHzwwQcAgHfffRcA8Nhjj/nrwZsvM5FIYG1tLXDOnjx5EslkEqVSCfPz8zh//rx//hHRnWeoKnlmXwhdp2Xb7cUXX0QikcDi4iKOHz/edlMpFou47777oJRCpVLB4uJiWyUtrunpadi2jcXFxUAb/t///vd45plnAK/A8tFHH+Hrr7+G67poNpt4/vnnta20i1NAGJQPP/wQAHDPPfeYi/CjH/0I0CoyvazbiWxHCu+6sLDNmpycxM2bN3Ho0CG/b5NZaIlSr9fRarVw7NixQJ+MPXv2BNaRStni4iImJib8c6LbPI6XL18GIiqdUkjppyK1WTKQhRS69uzZEzj+Y8eOAQC++eabwPfiGFQ+ksq9PPzZSjdu3DCDAK1yH8e///1vM6grx3HQarW6nheDitNe1Go1tFqttv5K8jvffvut+ZVYbsexxPGvf/3L/3+1WkWpVGo7dnlIF5VfupHKzd///nfUajW4rhtaGR4fH++pQtmv999/HwBw7733moswMTFhBg1MVN4JyxNxyYOaL774AtDKLWHxm0wmgZAHFefPnwcAnDp1yv8/Ed2ZhqqSt1uMjY3h+PHjgFfZMt+apdNpv+A9PT2NvXv39jQym0kqC/I203EcWJblF9DHx8dx7tw5jI2NYXx8HE8//XSgsDAo8qYl6uYYpdOxm0+ze1m3k16exg6KdMK///77gR6Gr5c3NeVyGV4T7LaPPEg4d+4clpaWcPPmTRw+fBhTU1Nd3950ygv626PbRdKq0Wi0HbdSqq9K3qDykUkG4TA/myHn1UcffRQIbzabSCQSm9rfQdqqOO2kXq8jmUy25Qn59PuW43YcS69c10Uul2s7Zvl0G/wqjm4PBbZjCoNugysN2vT0NCzLwvr6eiBc8oT51m0zzFYzOvNNopicnPTfMMubQSK6Mw1VJa/TTavTsu1Wr9dx6dIlZDIZuK6LP/7xj+YqAd988w0efvhhMzg2adohb/Peffdd/Pa3vzVX83377bd+05MoZvOnOKRZoHlzjCtsVEthNsvpZd2dQEb0W19fx9dff43Z2dnQJ9Nh5E3Nxx9/bC4KdfToUXz99dfI5/OoVqvYv39/rIKSNI8MY76N3k7y5Fyajw3SoPKRrHvffffBtu22z2ZMTk76b2hl5NtisYhqtYqzZ8+aq2+ZuHO6DSpO4/je976HarXa9iBtULbzWHoVVhEZJGkODQDXrl0LLDN9//vfN4N2tXPnzsF1Xb8rRa1Ww+LiInK53ECuheY2Oj2I09MBXvPV0dFRv9mmORo2Ed05hqqS99JLL8GyLDMYlmX5/QO2mjSh6FRozmazyOfzeO2115BMJnHq1KnI5iwLCwv49a9/3VelSidvhc6cOYPFxcXIp42O46Ber/v9YMKExXEcTz31FCzLguu6sZrPyvDrUggOK7BI4U36yfWybifSN67XvlT9OnHiBKrVKs6fPx/aNKeTyclJWJaFUqkUWpgNi+uxsTHMzs5iaWmpa586qZxfvXrVXITvvvsO8KbXuF1++tOfAgDefvttcxEQcfzdDCofmeum02lcuXKl7bNZr732GlKpFH7zm99gZGQEf/jDH1Aul0P7gg6aXJvW1tZCr3vS1HxQcdoLmcomKn/3kzdwm46lVwcPHkS1Wg2tINTr9ch7Tjfyxmrv3r3+W61WqxX6O/I22ay0DNJDDz0EaP3YtkM6nUYmk8EXX3yBkZERPPnkk3j55Zdx8uRJc9WemNdTeSP36aefBtaDkQ6iXq/j7NmzeO2113D+/HlYloVf/epXofcFIhp+Q1XJGx8fR7VaDXSGT6VSqFarW3qT0cmNPWywA3gVvCNHjvid+qXN/JEjR9ouxMViEd9++23HCldc6XQalmVhcXERmUwmtCJRq9Xw+uuv45133jEXAdrTRLPwUqvVYhUYxsbG/OZR2Ww2tFAgisWiX9GanJyEbdtotVpthTLpGyOV+F7W1Zlxv2/fPliWhWq12radTk21+hVWOJYbfifNZhPNZhPpdBqu62JmZiawrVqt5vepcxynrW9nt/540CrnpVKpbT/ff/99JBKJgb4pl2OKa3JyEolEAqVSqW1urGKx2HPT23q93nc+MtNM1o0a6CSKuZ1Omt5b4FOnTuGbb76BUgqff/75wCt4ZtrrZBCnVCrVFv9ynvcbp73kBZM8zDp27Fhb3s9ms21vy+P+Vr/Hshmd4j/ME088AXgVXf363PTm9pO+g71oegNS6W+sXn75ZSBkcK96vY5qtRra7FzP381m029S3esxQhuYZXFxse2eEjc9e5XNZvHoo4/iypUrfpPwQdynz549i2Qy6V9PX3jhBcDr1mEqlUrIZDJ+Oki65vN5f8CmN954A61WK7SPvTzQJaIhZg63SZsjw9SHDdu+tLTUNqS10obz1oeKl3nVhD7/TyKRUJZltQ3L303U0NHK22/5DeX9vrl9GT7bDA8bDroTGf7ZsiyVz+cD39vY2FCpVKptyOgNbx4kfTh0GSLcjOu460pcImKIe9lPWV6pVFRem/8slUr5cRE1L5RsX48zySN6nEnaJJNJ/3dkSG7bttXS0pI/hDy8oc0r3px78lv6vFC2batkMhkYal2G415aWgoMf24OGx5G4iJpzIWlx7G5rn58nUQdkz4nln7OyPxr+hDmkubw8pUMb29OvxBG9lfSU/JC3HykQtJPX9ccAj5Ko9EIHJu533p+0s8Pff/z+bz/qVQqbWkj29f3Sd+uflxR8RKmoU3Z0Cn+48Zpt3Ozn+OAl8ds21aWMSWInBuIuD6GiXssyjg/zaH3o+jXjlwup1zX9a8dlnHOyrliTnkg0xjAy5txh/nXr1FyvXC9uefMNNWPTY5b1jWPVbabTCZVuVz273OyX/o1tZc0luO0LEsVvDnuMpmMv24ul/OvRbJdPf17iVe5XiW9OUflUy6XVaVSaTtvw8i5kslklKvNbWkZU2MoY3qWhjYfo75PEmZOqdDwpnswj1fOiUTIFCBENDxYydsCtjehrk4KgTDmtpGbm/7R15WPbE/m0pGbbS82vAmjTWVvgmn996yQeZTkxmJKJpMq4U3EHJfruiqXy/mFA7nhyE0vzIZXAZT1paAQptu6eqFOPuYNUnlxIwWFhDfZthRKZHthaRi1/bC0zefzqqFNFCy/0/AKzgmv8iNkG7ZtBwoEcqOX7aZSqcDyije5rxxP2DY6KZfLgfQyt68i4gIxniWZxxS2nag4Ffr3pBAfp8CltP02v9MtHwnZf71gmUgkQisoUcxjgxZvYccu25ZrgrlcPlIoN8MREaf6PkfFSxjXdQNxFfWdbnEatk96OpvLEHEc+nfM81g/xqhzMo5ux6IijifOb8h1Wb/uhG0j7FzR5bWJ5cP2L4rrun7FAyHxptMrKd3WlYdDlvYAxLZtlfEmPFd9prF+nPp1JJPJ+NepsO/3Eq+yf/p10PyEVdRMjUZD5fN5P092u151W1ffVz3ezX2T+NrQJo8nouE1ov5zIaABqtVq2LNnD1zX3bJmotJE6/PPPzcXRVpeXsatW7f6albSbDZx//3347333tt0/0CiYbOwsIC5uTlUKpXbcn5ks1m/Segnn3zih9+6dQulUgnHjx+P1TSXiLpbWFjAgw8+iLvvvhvXr18PNOG/ePEiEolEWzNeIqLtxkreFikWi7h8+fKWXeiz2Swef/zxnvrdTExMYH19va+KpwyC0k8FkWjY3c5KXrdrwerqKv7v//6PlTyiASgWi/jb3/4WOchK05tndqvu/UREcQ3VwCs7STqdxhNPPIFsNmsu2pR6vY7l5WU8++yzkYU6nXSsdhwHBw8e7KuCVywW+34DSHQn2IrBeOKQods7ef/99/HUU0+ZwUTUh2739JWVFX/AFCKi24lv8raY4zi4cePGQEcfjCudTqNUKiGZTKLVavX1Fk/m2IlToSS6EzWbTezbtw+u62JpaWlb35jVvcm+W60WMpkMJicn/QnqP/nkE1y8eBHnz5/fMRN0E+12U1NTqFarsG0bBw4c8OewvX79Oi5duoQjR47clvs9EZGJlbwhJk3IbNtGPp9nQY9owBzHwSOPPBIIs217IHPfxVWv13HmzBmsr6/DdV3Am4g7lUrhxRdfDJ0uhYj602w2ceHCBVy8eBHVahXw5o5Np9M4evQo77NEtGOwkkdERERERDRE2CePiIiIiIhoiLCSR0RERERENERYySMiIiIiIhoirOQRERERERENEVbyiIiIiIiIhggreUREREREREOElTwiIiIiIqIhwkoeERERERHREGElj4iIiIiIaIiwkkdERERERDREWMkjIiIiIiIaIqzkERERERERDRFW8oiIiIiIiIYIK3lERERERERDhJU8IiIiIiKiIcJKHhERERER0RBhJY+IiIiIiGiIsJJHREREREQ0RFjJIyIiIiIiGiKs5BEREREREQ0RVvKIiIiIiIiGCCt5REREREREQ4SVPCIiIiIioiHCSh4REREREdEQYSWPiIiIiIhoiLCSR0RERERENERYySMiIiIiIhoirOQRERERERENEVbyiIiIiIiIhggreUREREREREOElTwiIiIiIqIhwkreFnMcB8vLy2bwrrG8vAzHcczgO1a9Xsf8/DxGR0cD8eI4DkZHRzE/Px9Y/05Wq9WQTqcxMjKCkZERzMzMoFarmatFKhaLmJmZwcLCgrloYOr1OhYWFtrS8043qLhfXV3F1NSUnwey2Syazaa5Gm0Sz7XtI3E1MzMTCE+n0xgdHUW9Xg+Eb4dms4lisYiJiYktTUMi2l1YydtCxWIRN27cwNGjR81Fu8bRo0dx48aNLam8yM1SCiajo6NIp9NwHAe1Wm3H3azq9TpWVlZQKpXQarXMxaSp1WrYv38/XnnlFSilUCgUcO3aNbzxxhvmqqGKxSIuX76MtbU1c9HANJtNrKys4PTp00xPzerq6kDivlgs4je/+Q2uXLmCRqOBfD6PxcVFrKysmKvSJvBc2z7FYhEfffTRlsZVP1ZWVvCHP/wBruuai4joTqZoSxQKBZVKpczgXSuVSql8Pm8G92VjY0MlEgllWZbK5/PKdV2llFKNRkMVCgWVSCQUgIH93qClUikFQFUqFXPRtmg0GmppackM3lFSqZRKJpNmcKSwY6pUKtuSD2zb3rL0jNr3qPCdYhBxb1mWyuVyZrBSu+D4dxOea9trY2NDAVC2bZuLbqt8Pr8taUhEuwff5G2BWq2Gw4cP49SpU+aiXetPf/oTTp8+velmNvLU+ebNm7h69SpmZ2cxPj4OABgbG0M6ncZnn32GZDKJv/71r+bXd4Sf/OQnZtC2unDhAm7dumUG7yjr6+sYGxszgyPthmPqVbPZxOnTp83gyPBhUqvV0Gq1cM8995iL4DjOjj23dyOea9trcnLSDCIi2pFYydsCc3NzSKVSfuVlGEgF7MiRI+ainjz33HNotVo4d+5c5M1ybGwMpVLJDCavudDc3JwZvOP00iRrtxxTL5rNJmZmZtriISp82Pz73/82gwCv8veLX/zCDKZN6CUvDeO5RkRE4VjJGzDHcbC2tob9+/cHwhcWFvy+Z/JxHKctfFDM7crvCb0vXNy+b48++ihc18Xq6mogfGpqKlaH89XVVVSrVViWhXQ6bS4OGB8fxy9/+ctAWLPZRDabxejoKEZGRjA1NYVisRhYR8RZVx9EBQDm5+cx4g0OIcztzMzM4Msvv9S28h9hAxeY219dXcXExIS/HXMAinq9Hhg8YWJiIjBoz8LCAg4fPgx4DxJkO8Lc16mpqbY3r/V63U976QOpH28n5uAO0n9SJ8sAYG1tzf/bXE90OyYhxyb7baYlIvavW540LS8v+2kU9f2weNb3p16vY9++fahWq4AWJ8ViMTRcj5uwYzDjrlarIZvN+vFk7rPkKxnkYmRkZKB9avXBVGSwIflNx3EwMjKCRx55BNDSdGRkBKurq9i/fz9arVYgb3RTr9fb4tu8BiFkv7LZbFv6OY7jx51UuCVtej1f4cWxrGOer81m008bx3FQLBYxOjqKqamp0G3p4uQDPf7u1HPNvM+FhevHqKe/rNftHDHzVdj90sw7Qr8vxIlX8x5gfqLS1uQ4jr/PYfcBYeZf2U9h/r7Emx4WFh9EtEOY7Td3O9d1/T5TAFQqlfL7fG2HTCYT2eegUqkoy7La2s3btq1SqZRqNBqB9TdraWnJj4eNjQ1zsUomk219MzqRfhuZTCYQ3uk3dJIu/fRlcF1XWZalMpmMajQaqtFo+HFt7k+cdRuNhsrn84H0yGQyyrIsZVmWv04ymVTJZFK5rhvYjp7G5XLZPzZJV3P7hUJB5fN5ValU/G3oeaDRaKhEIqGSyaS/z8lksi0vRfWdkfUlXPo9mumir+O6rrJtO1Z6lMtlBcDPL67r+vtXKBTM1XtK56hjkvBcLqcymYwql8uqXC77aaSfL3K85XJZKW1/E4lE1/NK+gnJeZjP5/24M78fJ2+Z2zVFhceJ442NDZXL5fz9zeVyfr6SPJjL5ULD+znXzTQpFAoqmUz6eUr2xex/HPV91WPe2NjYUJZl+fsu5wkAP62Vtx+WZfnnSsW71lqW5e+rGXepVMpPO9u2ezpflXetl+t2o9Hw01X2dWlpyU+/fD6vUqlU6DlpipMPdL3EZ1S67OZzTdaDcU6ZfefM9M/lcqpQKKhKpRJ6rVVeftfTQvKVvt2Gca0X+n0hTrw2Gg1lWZafp/R0N68tYfJenzzbtv3rvMQ3QvKcfm9TWnlB7kHK2yf9nickHvRzkIh2nvaSxi6mX+z1j2VZ21bRk4ty1M3OvCFubGwELqqDJhd58wIvN5ReSXzqNjY22m6OYTYzoIpt2yqRSJjB/jb13+91XX2fXNf180rGq/SZaRO2najCk2zfvBnCKJiFfV/yih4Wtp7ybvBmQU++LwVwKfSYBQvzeybJK2ZBw3VdPz+YcWQeXydRxyTh5vkhhRkzHc3vS4GuW+VG0shcTy+g6+v2mrdMYeG9xrGE6ed1o9EIDZdtxE0PFZEmso/muS7xpP9m2PdFL/tihzyEkPSXCo/8llkBkgK6+X0YD2k2Njb8uO3lfDXTRI9/oRe8ZR3zWqzrNR+okH3rJCpdhvVcC4sbiUe9TBAWLxLn5rFG5auwfZDt2rbdNV4lDsPOo7B4MMk2zQcuesXWXNcsF0U90JD0kXRLpVJt5xsR7TxD1VzzzJkzof0TWq0Wzpw5YwZvCWmKFdUR/tChQ8jlcnBdFydOnMCTTz6JUqkUuX6tVuupmYbp1VdfBbymMbqVlRVkMplAGLzmMd3m2jHjeHJyEtPT04GwMP0O71yv17G2toZUKmUuwuuvvw4AePPNN4Ee19XNzs4CXjPR8fFxNJtNLC4uIp1Ot6XNwYMHA3/HcejQITMoMAz3XXfdBcuy8OCDD/phd999t///bk6fPt3WvPUHP/gBAPj9G++66y4AwIkTJ/x1xsbG2r5n+vDDD9FqtfD4448HwsfHx5HJZNBqtbZ0WPynn366LQ0A4Pr164B3jqytrcG27cDy++67DwBw6dKlQHiUBx54IPD3+fPnAQAXL14ENpG34ugnjvfu3Rvo1ypxZIYPqm+w/L55ricSCcDIz4Mg6Wr2A56dnYVSCmmvyfe7774LAHjssccC66XTaSQSCaytraFmzBln27Z/HJOTk235q9v5+uabb+LgwYOB742NjcGyLLRarbbrtVyHx8bGIvsio898MEh3wrkG7xwJOy/0AYEuXLgAAHjqqae0Nf7z3V4dOHAgNF4/+eQT//9ffPEFYAzsInm0l3unOTDYyZMnYVkW1tbW/KaYb731FpLJZFscvPTSS4B3P9GVSiVYloVXXnkF2WwWiUTCP/+IaOcaqkpeWBt30WnZdnvxxReRSCSwuLiI48ePt11oRbPZxHPPPWcG92R6ehq2bWNxcTHQ5+H3v/89nnnmmcC68CoAcW4oZiFmK3344YcAEDpS349+9CNAq8j0sm4nsp0f/vCH5qLQsM2anJzEzZs3cejQITSbTSwsLLQVbqPU63W0Wi0cO3Ys0Fdiz549gXWkoLi4uIiJiQn/nOg2j+Ply5eBiEqnFEjiFu4GSUYI/OqrrwAAe/bsCRz/sWPHAADffPNN4HtxybHJg5tB5a0wOzWOdTJiptlPR47522+/Nb+yKVKpuvfee81FAZKPwwrRyWQS2IIKaLVaRalUaosLeQB248YN8yux7NR8MEznWlyyffP+bP49aGF9Ey3LMoN6IhXTr776CrVaDa7rhp4v4+PjoQ8qxsfH8cYbb6DVaqFYLOLkyZOB7xHRzjRUlbzdYmxsDMePHwe8ylZUJ/znn39+05U8AH5lQd5mOo4Dy7LanigvLCzg0UcfDYQNkjz97bUw2Gm4b/MYelm3k3/84x9m0JZrNpuYn5/H/fffD2hPrLv55z//CQAol8vwmmC3faRgcu7cOSwtLeHmzZs4fPgwpqam2t5ymP71r3+ZQT7zifztIGnVaDTajlsp1XfB0zSovBVmu+NYH3hJPmGDcOjq9TqSyWRb/Mrn3Llz5le2hdmyQGe+1RgU13WRy+Xa4kA+/b7l2O580KthONfiivOwc5CkRcW1a9f8MCkb9JufwkSNfCui3lQePXrUrwCaLYOIaGcaqkpepwthp2XbrV6v49KlS8hkMnBdF3/84x/NVTA/P4//+q//GsiNXZotydu8d999F7/97W8D66yuruLWrVux48lsshWH3MTW19fNRbGEjWoppMmY6GXdnUBG+ltfX8fXX3+N2dnZrm8whDTD/Pjjj81FoY4ePYqvv/4a+Xwe1WoV+/fvD316bJImW2G2+ul2J/K0/7PPPjMXDYSZX7Yyb21XHD/00EOwbTvweeihh8zVAr73ve+hWq1GPpTaKnqTtk46PayQ5oSDYllW39exOLYrH/RqmM61neapp55CMpnEq6++6uflEydOwLIsvxnlZn3/+9/3/69XJsPo68Irk7z88st+s8049wwiur2GqpL30ksvhTZrGORFshtpHtTpApjNZpHP5/Haa68hmUzi1KlTgaYRxWIR99xzT2i/kH7JW6EzZ85gcXEx0H+lVqvh7bffjt0EIyyO43jqqadgWRZc143VfFb6BcobwLBClRQ4pZ9cL+t2In3jpI/IVjtx4gSq1SrOnz8f2oymk8nJSViWhVKpFFoAD4vrsbExzM7OYmlpqWs/H6mcX7161VyE7777DvCm17hdfvrTnwIA3n77bXMREHH8cZj5ZVB5K8x2x/HJkydx5cqVwKfb+S/TwkTllX7jOcrDDz8MeP2HwsjbBOm39emnnxpr/P83QlFvJ/p18OBBVKvV0IplvV7vuzn7dueDXg3DuRaX3MvDpuvYCmNjYzh//jwmJiawf/9+jIyM4ObNm7h69eqmK/bNZhOJRALj4+OYnp7238iF5V99XbG6uopWq4XZ2VmcO3cOrVYr9tQ7RHT7DFUlb3x8HNVqNdBZO5VKoVqtbvoiGZfceKQJnSmbzeLIkSN+Z3/pcH7kyBH/5vXnP//Zn8dI5pt65JFH+i44wHubZ1kWFhcXkclkAhWJtbW1QP8SePMomYOvyA3BvLnWarVY+zY2NuY36cpms6E3GFEsFv2K1uTkJGzb9vsD6KTvhlTie1lXZ1aO9u3bB8uyUK1W27bTqSlRv8IeCkihrpNms4lms4l0Og3XdTEzMxPYVq1W8/v5OI7T1symW388aJXzUqnUtp/vv//+wDvhyzHFNTk5iUQigVKphGw2G/husVjsu+nthQsXAg+I+s1bwow7Ua/X+4rjXuJoEOTB0LFjx9ryUTabjf3m2WQer5ienkYikYDrum0Fymw2679VeuGFFwCv6bupVCohk8m0Xf83G3dPPPEE4FV89Wtf05sLTfqNiTjnMniubfpc0/dH7i/9NiF9+umnAQD/8z//E9huL/HVi1qthueeew5XrlzBzZs3oZRCsVjcdPNUx3FQrVZx9uxZP+zll18GQgZkq9frqFarga4CtVoNZ8+exWuvvQZ4ZYlUKoW1tbXQ78cpCxDRNjGH26TNkWHqzSGilTcPTdgQ2DI8sTl8tdKGUJahll3XVYlEQll9zFETNmxzmLAhlJU2JL/5u4iYhyeKDEFtWZbK5/OB721sbIQOz7zhzZelD+Euw5ibcR13XYnLqOOV/ZTllUpF5fN5P71SqZQfF1FzDMn29TiTPKLHmaRNMpn0f0eG5LZtWy0tLfnTO8AbUrvizeElvyX7Jd9JJpOB4fQlLy0tLfn7mPfmdzKH0jZJXCSNeZX0ODbX1Y+vk6hjkjg1hx+Xuaf0ocIlzeHlKxl+PeycMsn2ktr8b/Lbm8mH+nDkhULB31ZUeNw41vOPnm5R4ZLucdJZhOVnZaRtIpFQtm0rK2TIfznGRMjcafo5kcvlOu6T7Hvc38sYc6qZv69vz8y3vZyvSvtNiSc5XyU9G41GIK+a8RAlbj7Q1zX3LcqwnmuyjVQq5V8/5X4Fb566jY0NPy3Nc0G+r+cX/ZqaTCb9Oe4k3S3v/iX5TfKOfhy9xKtsV+a4zOfzamlpSVUqlVhpK79laWUDiS/zfNGPTeLS9ebl0/dJwsz7ox63+vGGxQER3T6s5G0B25voVSeFeLng6+tKuHx0UiiRm5zrzQWY9CYy7cWGN4ltN4io9EihyZRMJlUikehYWDO5rqtyuZx/o4F3g81kMpHb2fAqgLK+3HjDdFtX4lX/6OkiyuWyf+NKeBMASwVMtheWhlHb1/OBfKSgINuR32l4c2ZJgUzINmzbDtz8pWAr202lUoHlFW/SXzmesG10Ui6XA+llbl9FxAViPEsyjylsO1FxKvTvScGmW6FTFLxJvmW7YccmuuUt4WoPZPTzKSpcxYjjqPzTS7iel8JExb0wzwnzGMzvwkinsjcZtOTzbiTfQitYh8lrE2uHpX9YXMi2ovJWp+/INi2vwmPmA/N7iHEeiG75QEWkU5zfGMZzTRkPFqSSYVmW/yAhKi2jjl+FXFPlOCUfuq4bGledtmuGwUszveIU9jEramEqlUrs+JJjk/zb7VyWOAk7BvmebC8qPYloe42o/5zINEC1Wg179uyB67ptzYQGpekN1PH555+biyItLy/j1q1b/pxwvWg2m7j//vvx3nvv9TXoChEREYVrNps4ceIEnn32WXz33Xf4+9//7i/79ttvUSwWcfXq1U033ySiOwcreVukWCzi8uXLbf0JBiWbzeLxxx/vaXCWiYkJrK+v91XxlP55/VQQiYiIKJw8tL1y5UrkwFvz8/N45plnWMkjotiGauCVnSSdTuOJJ55oGzBgs+r1OpaXl/Hss8/GquBJ533HcXDw4MG+KnjFYrHvN4BEREQUbWVlxZ8IPkyz2USr1WIFj4h6wjd5W8xxHNy4cWOgI6LFlU6nUSqVkEwm0Wq1+nqLJ8NHx6lQEhERUW9WV1fx85//HJZlIZPJ4Mc//jHuvfdefPfdd/j444/xxRdf4J133ol8y0dEFIaVvCG2sLCAubk52LaNfD7Pp4BEREQ7kOM4ePPNN7G+vo5WqwV4c/U999xzsaa6ISIysZJHREREREQ0RNgnj4iIiIiIaIiwkkdERERERDREWMkjIiIiIiIaIqzkERERERERDRFW8oiIiIiIiIYIK3lERERERERDhJU8IiIiIiKiIcJKHhERERER0RD5f/geG0o9c2efAAAAAElFTkSuQmCC\"\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003eThe scaling factors were computed by dividing the known reference object dimensions (in cm) by the corresponding dimensions of the reference object in pixels. Fixed adjustments were also applied as offsets which were heuristically determined.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.4. Evaluation Metrics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe predicted egg dimensions (length and breadth) were compared with the ground truth values provided in the CSV file. Mean Absolute Error (MAE) was measured which is the average absolute difference between the actual and predicted values. Root Mean Square Error (RMSE) to measure the average magnitude of the error, with larger errors penalized more. Pearson Correlation Coefficient measuring the linear correlation between the actual and predicted dimensions was also derived. Additionally, the Mean Absolute Percentage Error (MAPE) was computed to assess the relative accuracy of the predictions for both egg length and breadth. The correlation between actual and predicted values was visualized through scatter plots and line graphs. Scatter plots are generated to visualize the relationships between actual and predicted egg dimensions. These plots help to visually assess the accuracy of the model. Z-scores are also computed for the actual dimensions (length and breadth) to identify any extreme outliers in the data.\u003c/p\u003e"},{"header":"3. Results and Discussion","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003cp\u003e\u003cstrong\u003e3.1. Training\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eThe Yolo training results indicated that both the training models converged similarly irrespective of the number of classes (egg for training model 1 and egg and reference for training model 2). The results demonstrated that both models performed well, achieving high precision and recall values, with mAP50 exceeding 0.97 and mAP50-95 reaching a maximum value of 0.965. This indicated strong detection capabilities for both models. However, Training Model 1 trained for fewer epochs, exhibited comparable performance while requiring less training time (10183.1s compared to 11107.3s for Training Model 2 (Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). The training losses for Model 1 were also slightly lower, and the validation losses for both models were nearly identical, although the final validation loss for Training Model 2 was marginally higher, suggesting a slight overfitting trend in the later epochs. Based on these findings, the testing results of Training Model 1 are represented in the paper. Figures (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e) depict the training results for the model. The combined F1 score for all classes peaked at 0.99 with the confidence threshold set at 0.861, indicating a balance between precision and recall. The models performed consistently across classes. The Precision-Confidence Curves, Recall-Confidence Curves, and Precision-Recall Curve show precision, recall remained high across a range of confidence thresholds with the area under the curve (AUC) approaching 1.0. The uniformity and balance in feature distributions indicated that the YOLO models were well-trained on the dataset ensuring generalizability and robust object detection performance. Model predictions during the training and validation are given in Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e. The shorter training time and lower training loss for Model 1 suggest that for applications where computational resources or time are constrained, a simpler class configuration may be preferred without sacrificing detection quality. In addition to using a YOLO algorithm similar to the present study, [\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e] incorporated a BiFPN to enhance small object detection accuracy and a CBAM to improve feature extraction, specifically targeting leaky eggs. In comparison, [\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e] suggests that YOLO-TLA demonstrated a 4.6% increase in [email protected] and a 4% increase in [email protected]:0.95 over the baseline YOLOv5s model, indicating improved detection capabilities for small objects.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eTraining Results for yolo11x OBB\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMetric\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTraining Model 2\u003c/p\u003e\n \u003cp\u003e(Last Epoch)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTraining Model 2\u003c/p\u003e\n \u003cp\u003e(Best Epoch)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTraining Model 1\u003c/p\u003e\n \u003cp\u003e(Last Epoch)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eModel 1 (Best Epoch)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEpochs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTime (s)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11107.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11107.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10183.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10183.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTrain Loss (Box/Cls/Dfl)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.305 / 0.202 / 1.576\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.305 / 0.202 / 1.576\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.295 / 0.213 / 1.608\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.295 / 0.213 / 1.608\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation Loss (Box/Cls/Dfl)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.294 / 0.217 / 1.511\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.210 / 0.192 / 1.670\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.210 / 0.192 / 1.670\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.210 / 0.192 / 1.670\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003emAP50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.977\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.978\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.978\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.978\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003emAP50-95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.941\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.965\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.965\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.965\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrecision (B)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.978\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRecall (B)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003cp\u003e\u003cstrong\u003e3.2. Testing\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eThe Z-scores values were mostly positive indicating that most of the egg predictions were larger than average in terms of their dimensions and therefore the offsets helped improve the predictions. The actual model predictions from the two testing models is given in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. The actual v/s predicted values for the two models and their correlations are given in Figs. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e and \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e. The testing results are given in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. The negative offsets of the models reveal an underestimation of actual measurements. Model 1\u0026apos;s smaller offsets suggest it may be more accurate in predicting both length and breadth dimensions. In this study, we also used Yolo object detection algorithms, which other researchers have reported to give good results. For example, compared to detection models such as faster R-CNN Yolo was reported to exhibit superior performance [\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e]. Detection and Identification of Horse Freeze Branding Using a Rotational YOLOv5 Deep Learning Model was achieved with an accuracy as high as 95.6% [\u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e]. Support Vector Machines (SVM) have also been reported to classify egg sizes into six categories, attaining an average accuracy of 87.58% [\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e]. Similarly, a study on hatchery egg classification using machine learning algorithms reported an accuracy of 80.4% [\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eTesting Results for the prediction of egg dimensions\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"3\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting Model 1\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting Model 2\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMean Absolute Error (Length)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.29\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMean Absolute Error (Breadth)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.26\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eRoot Mean Square Error (Length)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eRoot Mean Square Error (Breadth)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.32\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePearson Correlation (Length)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePearson Correlation (Length)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.81\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eLength Offset\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-1.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eBreadth Offset\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-1.5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eBuilding upon previous advancements in automatic egg grading systems, this study utilizes a YOLO-based model to predict the length and breadth of eggs with a high degree of precision. Unlike traditional methods that focus on broad categories like weight and size, our approach offers a more detailed understanding of egg morphology, which can be applied in breeding, quality control, and research applications. By leveraging a single camera, the cost of setting up this system remains low. However, the spherical nature of eggs introduces slight errors in the testing model, particularly in the 2D aspect. Despite these challenges, studies have demonstrated that models like Segformer achieve impressive results, with a mean intersection over union (IoU) of 96.15% and mean pixel accuracy of 97.17% for small batch egg image segmentation [\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e]. In contrast, YOLOv11 OBB models, which are designed to handle object rotations, excel in situations where the orientation of eggs varies, offering superior performance compared to traditional segmentation models. These models are also better equipped to handle complex backgrounds, a key advantage when dealing with images that do not have plain or simple backdrops. This added robustness enhances the overall detection accuracy, ensuring more reliable results even in environments with varied visual noise [\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e]. This added robustness enhances the overall detection accuracy, ensuring more reliable results even in environments with varied visual noise [\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e]. Furthermore, research by Chen et al. (2020) has shown that utilizing advanced deep learning architectures such as YOLO in agricultural applications can improve detection efficiency, enabling faster and more accurate measurements. Recent studies also highlight the utility of convolutional neural networks (CNNs) in improving segmentation accuracy, particularly in complex agricultural imaging tasks [\u003cspan class=\"CitationRef\"\u003e28\u003c/span\u003e].\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Conclusions","content":"\u003cp\u003eThis study highlights the significant potential of AI-driven computer vision techniques to accurately predict egg dimensions from 2D images. By automating the egg grading process in the poultry industry, we can revolutionize efficiency and productivity. The findings emphasize that integrating machine learning into egg size determination not only streamlines operations but also minimizes labor costs, making it an invaluable advancement for traditional grading practices.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eConflict of Interest Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that there is no conflict of interest\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo funding was received for this work\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHH conceived the experiment, AH, HH conducted formal analysis, AH, HH, PHB, and AM were involved with data preparation, AH, HH, wrote the manuscript, AH, HH, and AAK reviewed the manuscript, AH, HH, AAK, provided the resources\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eRoberts JR (2004) Factors Affecting Egg Internal Quality and Egg Shell Quality in Laying Hens. Jpn Poult Sci 41:161\u0026ndash;177. https://doi.org/10.2141/jpsa.41.161\u003c/li\u003e\n \u003cli\u003eZhu Y, Song D, Wu X, et al (2024) FEgg3D: A Non-Contact and Dynamic Measuring Device for Egg Shape Parameters and Weight Based on a Self-Designed Laser Scanner. Agriculture 14:1374. https://doi.org/10.3390/agriculture14081374\u003c/li\u003e\n \u003cli\u003eMatias Y (2023) Accelerating climate action with AI\u003c/li\u003e\n \u003cli\u003eWu R, He P, He Y, et al (2024) Egg production monitoring in commercial laying cages via the StrongSort-EGG tracking-by-detection model. Computers and Electronics in Agriculture 227:109508. https://doi.org/10.1016/j.compag.2024.109508\u003c/li\u003e\n \u003cli\u003eAlvarenga FAP, Borges I, Palkovič L, et al (2016) Using a three-axis accelerometer to identify and classify sheep behaviour at pasture. Applied Animal Behaviour Science 181:91\u0026ndash;99. https://doi.org/10.1016/j.applanim.2016.05.026\u003c/li\u003e\n \u003cli\u003eThangaraj JWV, Krishna NS, Devika S, et al (2024) Estimates of the burden of human rabies deaths and animal bites in India, 2022\u0026ndash;23: a community-based cross-sectional survey and probability decision-tree modelling study. The Lancet Infectious Diseases S1473309924004900. https://doi.org/10.1016/S1473-3099(24)00490-0\u003c/li\u003e\n \u003cli\u003eHamadani A (2024) Artificial Intelligence on Farms: Sheep Breed Classification Using Computer Vision. Indian Journal of Animal Production and Management 40:260\u0026ndash;268. https://doi.org/10.48165/ijapm.2024.40.4.8\u003c/li\u003e\n \u003cli\u003eHamadani A, Ganai NA, Khan NN, et al (2022) Comparison of various models for the estimation of heritability and breeding values\u003c/li\u003e\n \u003cli\u003eHamadani A, Ganai NA (2022) Development of a multi-use decision support system for scientific management and breeding of sheep. Sci Rep 12:19360. https://doi.org/10.1038/s41598-022-24091-y\u003c/li\u003e\n \u003cli\u003eHamadani A, Ganai NA (2023) Artificial intelligence algorithm comparison and ranking for weight prediction in sheep. Sci Rep 13:13242. https://doi.org/10.1038/s41598-023-40528-4\u003c/li\u003e\n \u003cli\u003eHamadani H, Khan A (2015) Automation in livestock farming \u0026ndash; A technological revolution. International Journal of Advanced Research 3:1335\u0026ndash;1344\u003c/li\u003e\n \u003cli\u003eYang Y, Wu L, Yin G, et al (2017) A Survey on Security and Privacy Issues in Internet-of-Things. IEEE Internet of Things Journal 4:1250\u0026ndash;1258. https://doi.org/10.1109/JIOT.2017.2694844\u003c/li\u003e\n \u003cli\u003eAb Nasir AF, Sabarudin SS, Abdul Majeed APP, Abdul Ghani AS (2018) Automated egg grading system using computer vision: Investigation on weight measure versus shape parameters. IOP Conf Ser: Mater Sci Eng 342:012003. https://doi.org/10.1088/1757-899X/342/1/012003\u003c/li\u003e\n \u003cli\u003eThipakorn J, Waranusast R, Riyamongkol P (2017) Egg weight prediction and egg size classification using image processing and machine learning. In: 2017 14th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON). IEEE, Phuket, pp 477\u0026ndash;480\u003c/li\u003e\n \u003cli\u003eUltralytics (2024) Oriented Bounding Boxes Object Detection\u003c/li\u003e\n \u003cli\u003ePython Software Foundation (2023) Python Language Reference, version 3.10.12\u003c/li\u003e\n \u003cli\u003eSonka M, Hlavac V, Boyle R (1993) Image pre-processing. In: Image Processing, Analysis and Machine Vision. Springer US, Boston, MA, pp 56\u0026ndash;111\u003c/li\u003e\n \u003cli\u003eSzabo FE (2015) E. In: The Linear Algebra Survival Guide. Elsevier, pp 101\u0026ndash;118\u003c/li\u003e\n \u003cli\u003eLuo Y, Huang Y, Wang Q, et al (2023) An improved YOLOv5 model: Application to leaky eggs detection. LWT 187:115313. https://doi.org/10.1016/j.lwt.2023.115313\u003c/li\u003e\n \u003cli\u003eJi C-L, Yu T, Gao P, et al (2024) Yolo-tla: An Efficient and Lightweight Small Object Detection Model based on YOLOv5. J Real-Time Image Proc 21:141. https://doi.org/10.1007/s11554-024-01519-4\u003c/li\u003e\n \u003cli\u003eJiang C, Li X, Ying Y, Ping J (2020) A multifunctional TENG yarn integrated into agrotextile for building intelligent agriculture. Nano Energy 74:104863. https://doi.org/10.1016/j.nanoen.2020.104863\u003c/li\u003e\n \u003cli\u003eHua Z, Jiao Y, Zhang T, et al (2024) Automatic location and recognition of horse freezing brand using rotational YOLOv5 deep learning network. Artificial Intelligence in Agriculture 14:21\u0026ndash;30. https://doi.org/10.1016/j.aiia.2024.10.003\u003c/li\u003e\n \u003cli\u003e\u0026Ccedil;elik A, Tekin E (2024) Classification of Hatchery Eggs Using a Machine Learning Algorithm Based on Image Processing Methods: A Comparative Study. Braz J Poult Sci 26:eRBCA-2023-1882. https://doi.org/10.1590/1806-9061-2023-1882\u003c/li\u003e\n \u003cli\u003eWaranusast R, Intayod P, Makhod D (2016) Egg size classification on Android mobile devices using image processing and machine learning. In: 2016 Fifth ICT International Student Project Conference (ICT-ISPC). IEEE, Nakhon Pathom, Thailand, pp 170\u0026ndash;173\u003c/li\u003e\n \u003cli\u003eLiu C, Wang Q, Ma M, et al (2023) Single-View Measurement Method for Egg Size Based on Small-Batch Images. Foods 12:936. https://doi.org/10.3390/foods12050936\u003c/li\u003e\n \u003cli\u003eKhanam R, Hussain M (2024) YOLOv11: An Overview of the Key Architectural Enhancements\u003c/li\u003e\n \u003cli\u003eLi J, Han Y, Guo L, Hao W (2022) Research on Lightweight Pedestrian Detection Model in Complex Background. In: 2022 International Conference on Machine Learning, Control, and Robotics (MLCR). pp 91\u0026ndash;95\u003c/li\u003e\n \u003cli\u003ePeng M, Liu Y, Qadri IA, et al (2024) Advanced image segmentation for precision agriculture using CNN-GAT fusion and fuzzy C-means clustering. Computers and Electronics in Agriculture 226:109431. https://doi.org/10.1016/j.compag.2024.109431\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Computer Vision, Egg Dimensions, Artificial Intelligence, YOLO","lastPublishedDoi":"10.21203/rs.3.rs-6016705/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6016705/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eEgg size determination is an important activity in the poultry industry. Traditional methods of size assessment are labour-intensive, error-prone, and time-consuming. Machine learning is proving to be a major game changer in all sectors and this has the potential to upgrade and automate egg grading as well. Considering all this, this research was undertaken to evaluate the potential of AI-driven computer vision approaches for the extraction of egg dimensions from 2D images. The images were annotated and saved as an OBB dataset and were appropriately preprocessed. Yolo11 OBB models were evaluated for their ability to detect eggs and yolo11x-obb model was found to be optimal for this study. The Mean Absolute Errors (MAE) of the two final predictive models for egg size were 0.28 for length and 0.19 for breadth, with Pearson Correlations of 0.81 for length and 0.85 for breadth for Model 1 and MAEs of 0.29 for length and 0.26 for breadth, with Pearson Correlations of 0.88 for length and 0.81 for breadth for Model 2.\u003c/p\u003e \u003cp\u003e \u003cb\u003eScience4ImpactStatement\u003c/b\u003e: This foundational research concludes that Machine Learning methods can predict egg dimensions with good accuracy and could potentially be used for poultry breeding, automatic egg grading, packaging, processing as well as marketing.\u003c/p\u003e","manuscriptTitle":"Enhancing Egg Grading Precision through AI and Computer Vision-Powered Morphometric Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-27 08:54:57","doi":"10.21203/rs.3.rs-6016705/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"291adf9b-77b0-41cd-abab-c125b9ba1324","owner":[],"postedDate":"February 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-02-27T08:54:57+00:00","versionOfRecord":[],"versionCreatedAt":"2025-02-27 08:54:57","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6016705","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6016705","identity":"rs-6016705","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}