A deep learning classification pipeline for identifying economically important tephritid fruit flies

A deep learning classification pipeline for identifying economically important tephritid fruit flies | 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 A deep learning classification pipeline for identifying economically important tephritid fruit flies Darren Ward, Diana Pérez-Staples, Christopher Weldon, Vijendra Samy, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7802967/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 Computer vision approaches utilising deep learning offer significant potential benefits for entomological applications, particularly for image-based taxonomic identification. Fruit flies (Tephritidae) represent economically damaging pests where species-level identification is critical for effective pest control, management, surveillance, and eradication programs. We assessed the capability of a deep learning convolutional neural network (CNN) pipeline to identify tephritid species from wing images, which serve as key diagnostic features. Our dataset comprised 1380 tephritid wing images spanning 34 tephritid species and 12 genera, with additional images from two 'other' classes (other Diptera and Hymenoptera). We employed a two-stage approach: (1) object detection using an Ultralytics YOLOv11n model to detect wing objects in images, treating all wings as a single class, followed by (2) species classification using an Ultralytics YOLOv11-cls model applied to cropped and augmented wing images generated from the object detection stage. The models demonstrated high accuracy in both wing detection (mAP50-95 value of 0.99 on a novel test set) and species classification (overall accuracy of 0.98 on a novel test set). Class-wise accuracy for different species varied (0.67-1) but showed general correlation with the number of original images available per class (10–285). Our results provide a potentially valuable tool for detecting pest tephritid species in biosecurity contexts. While deep learning technology remains in early development stages for entomological applications, such approaches hold promise to transform diagnostic and surveillance capabilities for biosecurity and pest management. Biosecurity Computer vision Insects Pest management Surveillance Tephritidae Figures Figure 1 Figure 2 Figure 3 Key Message Deep learning boosts insect ID through image-based computer vision. CNN accurately classifies 34 tephritid species from wing images. YOLOv11 models achieved 0.99 detection and 0.98 classification accuracy. Results aid biosecurity and pest management for fruit fly control. AI shows promise to transform entomological diagnostics and surveillance. Introduction Tephritidae, commonly known as fruit flies, are a diverse family of Diptera that are widely distributed around the world. Several species of Tephritidae cause severe economic damage to horticulture industries in a wide range of countries through costs associated with losses to fruit and vegetable crops, the loss of market access, and the costs associated with control (Suckling et al. 2014; Dias et al. 2018 ). Such losses can be substantial, for example, He et al. ( 2023 ) estimated the economic losses in China from Zeugodacus cucurbitae at many hundreds of millions of dollars per year. Kibira et al. ( 2010 ) noted that in eastern and southern Africa, several species of Ceratitis contribute to mango yield losses of between 30% to 70% depending on the locality, variety and season, and that an exotic species, Bactrocera dorsalis (reported as B. invadens ) causes damage to over 80% of mango crops. In Australia, costs associated with maintaining exclusion zones and periodic outbreaks of B. dorsalis (reported as B. papayae ) can also be as high as hundreds of millions of dollars (Jessup et al. 2007 ; Suckling et al. 2014). Considerable effort and expenditure are associated with different types of control and management (Dias et al. 2018 ; He et al. 2023 ), such as the use of lures (Stringer et al. 2019), area wide suppression (Lloyd et al. 2010), natural enemies (Dias et al. 2022 ), and the sterile insect technique (Shelly and McInnis 2016 ). However, there is also significant effort and investment in keeping countries and areas free of these pests, particularly as many species can be easily transported around the globe via human trade networks. Such investment includes quarantine and fruit destruction, surveillance networks via large scale trapping, and eradication programmes (Quilici and Donner 2012; Suckling et al. 2014; Kean et al. 2024 ). Correct species identification is critical, whether it is for the control and management of pests, or for keeping areas pest-free through surveillance and eradication. However, the morphological identification of tephritid species can be difficult and requires practical experience and professional skills (White 2006 ; Plant Health Australia 2011 ; Drew and Romig 2013; He et al. 2023 ). Molecular techniques are increasingly being used to resolve species limits and distinguish species, but this is also not without problems (Dhami et al., 2016 ; Doorenweerd et al. 2023; Starkie et al., 2024). Imaged-based identification offers an additional, and likely complementary, approach to morphological and molecular identification. Computer vision approaches based on machine learning and deep learning have significantly influenced a wide range of scientific disciplines, though these have only relatively recently been applied to entomology (Boer and Vos 2018 ; Marques et al. 2018; Hansen et al. 2019). Recent studies on image-based insect identification are showing that deep learning models can extract features from images and learn to differentiate species to an accuracy approaching, or exceeding, human expertise (Valan et al. 2019 ; Hoye et al. 2020, Ward and Martin 2023 , da Cunha et al. 2024 ) and is being applied to a growing range of insect taxa. For example, over half of British ground beetles (Carabidae) can be identified to species, and 74% to genus using convolutional neural networks (Hansen et al. 2019). Boer and Vos ( 2018 ) used over 10,000 images from AntWeb ( www.antweb.org ) to classify ants at species level based on dorsal, head, and profile images, obtaining accuracy between 62–92% for species and 79–95% for genus, depending on different configurations of the models. Previous work on the automated identification of fruit flies is limited, however Wang et al. (2016) used custom built software (based on C# language on Microsoft’s Net Framework) to classify 74 species belonging to six genera, and despite having only a few images per species was able to obtain good identification rates depending on the morphological character, 71% (wing), 69% (thorax), 39% (abdomen). More recent work includes Tariq et al. ( 2022 ) who were testing the development of identification systems for farmers and growers through mobile applications and examined the image capture of two species using smartphone and Raspberry Pi cameras in both the laboratory and fruit orchards. In this paper, we test the ability of a deep learning convolutional neural network pipeline to classify genera and species of Tephritidae using images of the wing for a range of species that are relevant around the world for pest control and management, and for border biosecurity and surveillance. Methods Taxonomic species and specimens We captured 1,380 wing images representing 34 species across 12 genera of Tephritidae, along with 1,151 images of other Diptera and 3,368 images of Hymenoptera, totalling 5,899 images. An additional 10 tephritid species were excluded from the dataset as each had fewer than 10 images available. The dataset included economically important species from the genera Bactrocera , Ceratitis , and Zeugodacus , native New Zealand species from Austrotephritis , Sphenella , and Trupanea , and biocontrol agents from Procecidochares and Urophora that have been introduced to New Zealand for invasive weed control. Specimens were primarily obtained from the New Zealand Arthropod Collection, including species from New Zealand and Pacific Island nations, using dry pinned material. Additional specimens were obtained from laboratory colonies kept at the USDA Daniel K. Inouye U.S. Pacific Basin Agricultural Research Center in Hilo, Hawaii ( B. dorsalis, Bactrocera latifrons, and Z. cucurbitae ), the Indian Institute of Horticultural Research in Bangalore, India ( Bactrocera tau ) and South Africa ( B. dorsalis and Ceratitis cosyra ). The median number of images per tephritid species was 19 (range: 10–285; see Table 1 ). Table 1 Classification performance by species. Species Training Images Validation Images Test Images Accuracy Precision Recall F1 Score Austrotephritis cassiniae 32 6 2 1.00 1.00 1.00 1.00 Austrotephritis plebeia 44 8 4 1.00 1.00 1.00 1.00 Bactrocera caliginosa 38 7 3 1.00 1.00 1.00 1.00 Bactrocera correcta 57 10 5 1.00 1.00 1.00 1.00 Bactrocera curvipennis 38 7 3 1.00 1.00 1.00 1.00 Bactrocera distincta 176 33 11 1.00 1.00 1.00 1.00 Bactrocera dorsalis 912 171 57 1.00 0.97 1.00 0.98 Bactrocera facialis 313 58 21 1.00 0.95 1.00 0.98 Bactrocera frauenfeldi 48 9 3 1.00 1.00 1.00 1.00 Bactrocera kirki 345 64 23 0.96 1.00 0.96 0.98 Bactrocera latifrons 96 18 6 0.67 0.80 0.67 0.73 Bactrocera melanotus 124 23 9 1.00 1.00 1.00 1.00 Bactrocera minuta 83 15 6 0.67 1.00 0.67 0.80 Bactrocera obscura 89 16 7 1.00 1.00 1.00 1.00 Bactrocera passiflorae 320 60 20 1.00 0.91 1.00 0.95 Bactrocera tau 57 10 5 0.80 1.00 0.80 0.89 Bactrocera tryoni 57 10 5 1.00 1.00 1.00 1.00 Bactrocera umbrosus 131 24 9 1.00 1.00 1.00 1.00 Bactrocera xanthodes 211 39 14 1.00 1.00 1.00 1.00 Ceratitis capitata 64 12 4 1.00 1.00 1.00 1.00 Ceratitis cosyra 249 46 17 1.00 1.00 1.00 1.00 Coelotrypes punctilabris 70 13 5 1.00 1.00 1.00 1.00 Dacus solomonensis 44 8 4 1.00 1.00 1.00 1.00 Dacus vertebratus 185 34 13 1.00 1.00 1.00 1.00 Diptera 912 171 57 1.00 1.00 1.00 1.00 Dirioxa pornia 76 14 6 1.00 1.00 1.00 1.00 Hymenoptera 912 171 57 1.00 1.00 1.00 1.00 Procecidochares alani 32 6 2 1.00 1.00 1.00 1.00 Procecidochares utilis 41 7 4 1.00 1.00 1.00 1.00 Sphenella fascigera 44 8 4 0.75 1.00 0.75 0.86 Sphenella ruficeps 51 9 4 1.00 0.80 1.00 0.89 Trupanea fenwicki 41 7 4 1.00 1.00 1.00 1.00 Trupanea longipennis 38 7 3 1.00 1.00 1.00 1.00 Urophora cardui 32 6 2 1.00 1.00 1.00 1.00 Urophora stylata 32 6 2 1.00 1.00 1.00 1.00 Zeugodacus cucurbitae 233 43 16 1.00 1.00 1.00 1.00 Specimen preparation and imaging For pinned specimens, wings were carefully removed from the body prior to imaging. Specimens were placed in a specimen manipulator, and a micropin or number 5 forceps was used to gently manipulate the tegula until the wing detached naturally. Wings were not forcibly pulled to avoid membrane damage. Wings were positioned on a microscope stage with a white background, imaged, and subsequently stored in gelatin capsules for preservation. For specimens from Hawaiian, Indian and South African laboratory colonies, wings were secured in position using double-sided tape during imaging. All wing images were captured using a Leica EZ4 W microscope (Leica Microsystems) with an integrated camera. The wings provided from South Africa were prepared as reported in Roets et al. ( 2018 ) and Makumbe et al. ( 2020 ). Wings were positioned flat, eliminating the need for image stacking techniques (Fig. 1 ). Model training and validation Wing images were uploaded to Roboflow ( https://roboflow.com/ ), an online platform for image annotation and deep learning. A project was established in Roboflow (publicly available at: https://universe.roboflow.com/bugider/tephritidae ), where polygons were drawn around each wing using the Smart Polygon tool powered by Meta AI's Segment Anything Model (SAM2). Each polygon was annotated with the corresponding taxonomic species designation. Images were exported to a local workstation (HP Z2 SFF G9 with 128GB RAM and NVIDIA RTX A4000 GPU, Windows 10) for processing. Annotations were converted to bounding box format compatible with YOLO object detection models. Images underwent preprocessing including resizing to 640 × 640 pixels, grayscale conversion (due to limited colour information in tephritid wings), and exposure correction using contrast stretching to create the object detection training dataset. We used YOLOv11 models from the Ultralytics Python library (Jocher et al., 2023 ). We trained a two-stage model, first with a detection stage (finding the wing in the image) and then a classification stage (identification to species). A single object detection model was trained treating all wings as a single class. Multiple classification models were trained with varying dropout rates (0.2–0.5) and learning rates (0.0001–0.01) to optimize performance. Complete training parameters are detailed in Supplementary Table 1. Both datasets (detection and classification) were partitioned into training (85%), validation (15%), and test (5%) splits using stratified sampling to maintain proportional class representation across splits. For the classification dataset (i.e. stage two cropped wings generated from the detection stage), images were further processed by augmenting through random flipping, blurring, and noise addition. The two 'other' classes (Hymenoptera and Diptera) were randomly subsampled to match the size of the largest tephritid class, preventing training bias. The final classification dataset comprised 7,800 images. Model performance was evaluated on the test splits in each image set (unseen during training or validation) using mean Average Precision (mAP) for object detection and accuracy, precision, recall, and F1 score for species classification. We also assessed performance by comparing the confidence scores of correct versus incorrect predictions on the test set. Results and Discussion Both the detection and classification models converged well after 30 training epochs. Training and validation loss curves were similar for both models (Supplementary Fig. 1). Object detection performance The model (YOLOv11n) demonstrated excellent performance in detecting wing objects within images, achieving precision and recall of 1.00 for both metrics on the test dataset. The mean Average Precision at IoU threshold 0.5 (mAP50) was 1.00 and at mAP50-95 was 0.99. These results indicate that the model was highly effective at locating and delineating wing boundaries across all images, successfully distinguishing wing objects from the background. While these results may appear unrealistically high, it should be noted that wings were always displayed on a plain background, there was a single wing per image, and all classes were lumped for the object detection task, so precision and recall were expected to be very high. Further, we would expect that in any deployment of this model (i.e. for identification queries) the images would be captured in very similar conditions to the training images to ensure similarly high detection rates (see Methods). Images based from specimens photographed in field situations or of the whole body (wings not removed) are not expected to perform as well. Species classification performance The classification model (YOLOv11n-cls) showed strong overall performance across the 34 tephritid species and two other taxonomic groups. Overall (across classes) classification accuracy was 0.98, with precision of 0.98, recall of 0.97, and F1 score of 0.97, demonstrating robust species-level identification. Classification performance varied considerably among species, revealing patterns related to dataset composition and morphological distinctiveness (Table 1 ). Twenty-six species (76%) achieved perfect classification metrics (accuracy, precision, recall, and F1 scores of 1.00), including economically important pest species such as B. dorsalis , B. tryoni , C. capitata , and Z. cucurbitae . Several species showed slight reductions in precision while maintaining perfect recall and accuracy. Bactrocera passiflorae achieved 0.91 precision, B. facialis recorded 0.95 precision, and Sphenella ruficeps showed 0.80 precision. These lower precision values suggest occasional misclassification of other species (Fig. 2 , Supplementary Fig. 2). The most challenging classifications involved species with smaller training datasets and that were visually like other species in the same genus. Bactrocera minuta (83 training images) achieved 67% accuracy with precision of 1.00 but recall of 0.67, indicating that while wings predicted as B. minuta were accurately classified, some individuals within this class were misassigned to the visually similar B. passiflorae (e.g. Figure 3 a). Similarly, B. latifrons (96 training images) showed 67% accuracy with 0.80 precision and 0.67 recall (misclassified as B. dorsalis , e.g. Figure 3 b), B. tau (57 training images) achieved 80% accuracy with precision of 1.0 and 0.8 recall (misclassified as B. latifrons ), and Sphenella fascigera (44 training images) achieved 75% accuracy with precision of 1.0 and 0.75 recall (misclassified as S. ruificeps , e.g. Figure 3 c). Bactrocera kirki , despite having a larger training set (345 images), showed slightly reduced performance (96% accuracy, 1.00 precision, 0.96 recall, misclassified as B. fascialis ), suggesting morphological similarity to closely related species rather than sample size limitations. All misclassified images were assigned to other species within the same genus as the source species. Wing morphology is also much more conserved within genera than across genera. The high performance across diverse genera, including functionally distinct groups such as agricultural pests (e.g. Bactrocera , Ceratitis ), biocontrol agents ( Procecidochares , Urophora ) and native New Zealand species ( Austrotephritis , Trupanea ), demonstrates the classification model's ability to generalise across varied wing morphologies to accurately identify species of economic importance. Further, there was generally good separation in confidence scores between correct and incorrect predictions, with correct predictions skewed towards 1.0 and incorrect predictions typically below 0.8 (Fig. 3 ). Conclusions In this study we demonstrated the successful application of deep learning methods for automated species-level identification of Tephritidae, achieving excellent accuracy (98%) in distinguishing among 34 fruit fly species using wing morphology. The two-stage YOLOv11 pipeline showed perfect object detection performance and robust classification across economically critical pest species including B. dorsalis , B. tryoni , C. capitata , and Z. cucurbitae , which achieved near-perfect identification rates. Our results represent a significant advancement for biosecurity and pest management applications, where rapid and accurate species identification is essential for effective quarantine decisions, surveillance programs, and targeted control strategies. The classification model's ability to reliably distinguish between pest species, beneficial biocontrol agents, and native taxa addresses a critical bottleneck in entomological diagnostics, particularly in resource-limited settings where taxonomic expertise may be unavailable. By providing an accessible, automated identification tool that maintains high accuracy across diverse geographic specimens, this technology has the potential to transform frontline biosecurity operations. The further development of desktop or mobile applications will also enable real-time pest detection at ports of entry and support management decisions that protect agricultural systems and native ecosystems from invasive species. Declarations Competing Interests The authors have no relevant financial or non-financial interests to disclose. Funding This work was supported by SSIF Infrastructure funding for Nationally Significant Collections and Databases at Manaaki Whenua-Landcare Research; the B3 (Better Border Biosecurity) project D22.15 on ‘Using images and deep learning for the identification of high-risk insect species’; and the University of Auckland Summer Research program for Tomas Blokker, Angie Zhu, and Hope Ryu. Author Contribution DW and AH conceived and designed research. DW, AH, CW, VS, and DPS provided images. DB, KW and BM provided computer science support. DW and AH wrote the manuscript. All authors read and approved the manuscript. Acknowledgement Part of the images was made possible thanks to the Cooperative Agreements (23-8130-1011-IA, 25-8130-1011-CA) from the United States Department of Agriculture’s Animal and Plant Health Inspection Service (APHIS) awarded to DPS. It may not necessarily express APHIS’ views. DPS thanks Abraham Moreno Mejía and Kamala Jayanthi. Thanks to Tomas Blokker, Angie Zhu, and Hope Ryu at the University of Auckland for taking images. Data Availability Specimen records are accessible through GBIF (https://www.gbif.org) or Symbiota's Ecdysis portal (https://ecdysis.org). Wing images are available via Roboflow (https://universe.roboflow.com/bugider/tephritidae), and Python scripts are provided on GitHub ( [https://github.com/aharmer/tephritID](https:/github.com/aharmer/tephritID) https://github.com/aharmer/tephritID). References Ärje J, Melvad C, Jeppesen MR, Madsen SA, Raitoharju J, Rasmussen MS, Iosifidis A, Tirronen V, Gabbouj M, Meissner K, Høye TT (2020). Automatic image-based identification and biomass estimation of invertebrates. Methods in Ecology and Evolution 11:922–931. Bjerge K, Nielsen JB, Sepstrup MV, Helsing-Nielsen F, Høye TT (2021). An automated light trap to monitor moths (Lepidoptera) using computer vision-based tracking and deep learning. Sensors 21:343. Boer MJ, Vos RA (2018). Taxonomic classification of ants (Formicidae) from images using deep learning. bioRxiv 407452. da Cunha VAG, Pullock DA, Ali M, Neto AdOC, Ampatzidis Y, Weldon CW, Kruger K, Manrakhan A, Qureshi, J (2024). Psyllid Detector: A web-based application to automate insect detection utilizing image processing and deep learning. Applied Engineering in Agriculture 40:427–438. Dhami MK, Gunawardana DN, Voice D, Kumarasinghe L (2016). A real-time PCR toolbox for accurate identification of invasive fruit fly species. Journal of Applied Entomology 140:536–552. https://doi.org/10.1111/jen.12286 Dias NP, Montoya P, Nava DE (2022). A 30-year systematic review reveals success in tephritid fruit fly biological control research. Entomologia Experimentalis et Applicata 170:370–384. https://doi.org/10.1111/eea.13157 Dias NP, Zotti MJ, Montoya P, Carvalho IR, Nava DE (2018). Crop protection, fruit fly management research: a systematic review of monitoring and control tactics in the world. Crop Protection 112:187–200. Doorenweerd C, Leblanc L, Norrbom AL, Jose MS, Rubinoff D (2018). A global checklist of the 932 fruit fly species in the tribe Dacini (Diptera, Tephritidae). Zookeys 730:19–56. Drew RAI, Romig M (2013). Tropical Fruit Flies of South-East Asia (Tephritidae: Dacinae): Indomalaya to North-West Australasia. CABI, Wallingford. https://doi.org/10.1079/9781780640358.0000 Greeff M, Caspers M, Kalkman V, Willemse L, Sunderland BD, Bánki O, Hogeweg L (2022). Sharing taxonomic expertise between natural history collections using image recognition. Research Ideas and Outcomes 8:e79187. Hansen OL, Svenning JC, Olsen K, Dupont S, Garner BH, Iosifidis A, Price BW, Høye TT (2020). Species‐level image classification with convolutional neural network enables insect identification from habitus images. Ecology and Evolution 10:737–747. He Y, Xu Y, Chen X (2023). Biology, ecology and management of tephritid fruit flies in China: a review. Insects 14:196. https://doi.org/10.3390/insects14020196 Høye TT, Ärje J, Bjerge K, Hansen OLP, Iosifidis A, Leese F, Mann HMR, Meissner K, Melvad C, Raitoharju J (2021). Deep learning and computer vision will transform entomology. Proceedings of the National Academy of Sciences 118(2):e2002545117. Jessup AJ, Dominiak B, Woods B, De Lima CPF, Tomkins A, Smallridge CJ (2007). Area-Wide Management of Fruit Flies in Australia. In: Vreysen, M.J.B., Robinson, A.S., Hendrichs, J. (eds) Area-Wide Control of Insect Pests. Springer, Dordrecht. https://doi.org/10.1007/978-1-4020-6059-5_63 Jocher G, Qiu J, Chaurasia, A (2023). Ultralytics YOLO (Version 8.0.0) [Computer software]. https://github.com/ultralytics/ultralytics Kean JM, Manoukis NC, Dominiak BC (2024). Review of surveillance systems for tephritid fruit fly threats in Australia, New Zealand, and the United States. Journal of Economic Entomology 117:8–23. https://doi.org/10.1093/jee/toad228 Kibira M, Affognon H, Njehia B, Muriithi B, Mohamed S, Ekesi S (2010). Economic evaluation of integrated management of fruit fly in mango production in Embu County, Kenya. African Journal of Agricultural and Resource Economics 10:343–353. Makumbe LDM, Moropa TP, Manrakhan A, Weldon CW (2020). Effect of sex, age and morphological traits on tethered flight of Bactrocera dorsalis (Hendel) (Diptera: Tephritidae) at different temperatures. Physiological Entomology 45:110-119. Plant Health Australia (2011). The Australian Handbook for the Identification of Fruit Flies. Version 1.0. Plant Health Australia. Canberra, ACT. Roets PD, Bosua H, Ruth Archer C, Weldon CW (2018). Life-history and demographic traits of the marula fruit fly Ceratitis cosyra : potential consequences of host specialization. Physiological Entomology 43:259-267. Shelly T, McInnis D (2016). Sterile insect technique and control of tephritid fruit flies: do species with complex courtship require higher overflooding ratios? Annals of the Entomological Society of America 109:1–11. https://doi.org/10.1093/aesa/sav101 Tariq S, Hakim A, Siddiqi AA, Owais M (2022). An image dataset of fruitfly species (Bactrocera zonata and Bactrocera dorsalis) and automated species classification through object detection. Data Brief 43:108366. https://doi.org/10.1016/j.dib.2022.108366 Wang J, Chen X, Hou X, Zhou L, Zhua C, Jia L (2017). Construction, implementation and testing of an image identification system using computer vision methods for fruit flies with economic importance (Diptera: Tephritidae). Pest Management Science 73:1511–1528. https://doi.org/10.1002/ps.4487 Ward DF, Martin B (2023). Trialling a convolution neural network for the identification of Braconidae in New Zealand. Journal of Hymenoptera Research 95:95–101. White IM (2006). Taxonomy of the Dacina (Diptera: Tephritidae) of Africa and the Middle East. African Entomology, Memoir No. 2, 156 pp. Valan M, Makonyi K, Maki A, Vondrácek D, Ronquist F (2019). Automated taxonomic identification of insects with expert-level accuracy using effective feature transfer from convolutional networks. Systematic Biology 68:876–895. Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx 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-7802967","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":534780020,"identity":"7d98f068-bf10-4342-8a1e-69600f5c7ca2","order_by":0,"name":"Darren Ward","email":"data:image/png;base64,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","orcid":"","institution":"Manaaki Whenua – Landcare Research","correspondingAuthor":true,"prefix":"","firstName":"Darren","middleName":"","lastName":"Ward","suffix":""},{"id":534780021,"identity":"b6c97ce9-6381-4901-8900-da0ea8d84f25","order_by":1,"name":"Diana Pérez-Staples","email":"","orcid":"","institution":"Universidad Veracruzana","correspondingAuthor":false,"prefix":"","firstName":"Diana","middleName":"","lastName":"Pérez-Staples","suffix":""},{"id":534780022,"identity":"6db86ea9-08b2-41e2-9d08-17b8e0d3cbf8","order_by":2,"name":"Christopher Weldon","email":"","orcid":"","institution":"University of Pretoria","correspondingAuthor":false,"prefix":"","firstName":"Christopher","middleName":"","lastName":"Weldon","suffix":""},{"id":534780023,"identity":"903c6e35-1b02-487d-99ac-49c945b8e7ab","order_by":3,"name":"Vijendra Samy","email":"","orcid":"","institution":"Ministry of Agriculture and Waterways","correspondingAuthor":false,"prefix":"","firstName":"Vijendra","middleName":"","lastName":"Samy","suffix":""},{"id":534780024,"identity":"f12fddb6-d2cc-40ff-8b32-9df5997289fe","order_by":4,"name":"Daniel Bentall","email":"","orcid":"","institution":"Plant \u0026 Food Research","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"","lastName":"Bentall","suffix":""},{"id":534780025,"identity":"ac21b510-94fb-4054-82b3-6be597ac60ca","order_by":5,"name":"Brent Martin","email":"","orcid":"","institution":"Manaaki Whenua – Landcare Research","correspondingAuthor":false,"prefix":"","firstName":"Brent","middleName":"","lastName":"Martin","suffix":""},{"id":534780026,"identity":"9be97502-ea6a-4a91-bd28-280928a050d6","order_by":6,"name":"Katarina Taskova","email":"","orcid":"","institution":"University of Auckland","correspondingAuthor":false,"prefix":"","firstName":"Katarina","middleName":"","lastName":"Taskova","suffix":""},{"id":534780027,"identity":"e3b8ceff-b574-4a23-964b-91fc0226eb6c","order_by":7,"name":"Aaron Harmer","email":"","orcid":"","institution":"Manaaki Whenua – Landcare Research","correspondingAuthor":false,"prefix":"","firstName":"Aaron","middleName":"","lastName":"Harmer","suffix":""}],"badges":[],"createdAt":"2025-10-07 23:53:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7802967/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7802967/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94663728,"identity":"f5de7175-e6db-42f5-b069-0fcc938447a5","added_by":"auto","created_at":"2025-10-29 12:13:11","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":924984,"visible":true,"origin":"","legend":"","description":"","filename":"WardetalDeeplearningfruitfly.docx","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/dd269ce82fb2916fe344e472.docx"},{"id":94672611,"identity":"b7b82768-89c0-4c28-bee1-a3f48fff7d15","added_by":"auto","created_at":"2025-10-29 13:40:45","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9689,"visible":true,"origin":"","legend":"","description":"","filename":"03a8449c51f44934bace853f4555d271.json","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/7ae526f0f4598c0d55403694.json"},{"id":94663729,"identity":"fd93fd12-dbc3-4116-8f51-31fa486284c4","added_by":"auto","created_at":"2025-10-29 12:13:11","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1116728,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/1c3c8ee8cf277d856d9525c9.docx"},{"id":94673118,"identity":"f32476ae-8cea-4ed1-8768-78def298d88a","added_by":"auto","created_at":"2025-10-29 13:41:13","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":100623,"visible":true,"origin":"","legend":"","description":"","filename":"03a8449c51f44934bace853f4555d2711enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/618af40eb191cc3e7c714a28.xml"},{"id":94672553,"identity":"ae8af24b-ec37-4bb2-a814-660da6bcdb14","added_by":"auto","created_at":"2025-10-29 13:40:42","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1082212,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/f20bd9a041a870bc93eb7ee6.jpeg"},{"id":94672683,"identity":"d52ef4d3-b2c2-4dd5-909d-0dcf876ff872","added_by":"auto","created_at":"2025-10-29 13:40:50","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":68042,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/3da5583ddba4d87b3bd1118b.jpeg"},{"id":94663732,"identity":"95905fce-817b-4597-85d9-74a1e779b878","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":86140,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/c0227cb9df0901eb99993cee.jpeg"},{"id":94663742,"identity":"8a4e2c68-9c1f-4fb2-8849-d23a95167187","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":113592,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/da4bcd390b51543b353f11fa.jpeg"},{"id":94663740,"identity":"6e17796e-5ffe-4169-9c94-fdf2e33d3d7c","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":79471,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/b093e3867adfc5b3bb83d324.jpeg"},{"id":94673069,"identity":"ac076b7f-2569-4679-b88b-b67c699e6301","added_by":"auto","created_at":"2025-10-29 13:41:12","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108088,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/a9d03799d1f5bef4ad775ce1.jpeg"},{"id":94672707,"identity":"3604030f-c6f6-48b1-ab56-b3802e9c824d","added_by":"auto","created_at":"2025-10-29 13:40:51","extension":"jpeg","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":67476,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/ddd5738023fb4ba99fb0a3b2.jpeg"},{"id":94673419,"identity":"6f2d6d57-d095-45be-8b6c-2b582312b2dd","added_by":"auto","created_at":"2025-10-29 13:41:23","extension":"jpeg","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30641,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/871d30c25557f823b8f13e06.jpeg"},{"id":94663746,"identity":"fc0b8f7d-2f54-49a8-bccb-855233d17db4","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":286307,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/650380d083d4f5b2930ba5b1.png"},{"id":94663748,"identity":"4327a365-0b23-4786-b1be-a31e75ca543a","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149648,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/6e95fae9d8ff4581d5915255.png"},{"id":94663753,"identity":"58b53022-a039-406a-b101-fda0951725ef","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":154044,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/c93f1e46754b0f74a5aede04.png"},{"id":94672546,"identity":"0c9f7b95-788e-47cb-b582-6478f5006652","added_by":"auto","created_at":"2025-10-29 13:40:42","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":101746,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/933557c4671afbfb6958a688.png"},{"id":94673263,"identity":"b012dcc2-84a9-4a8d-84d3-ada29a8b7543","added_by":"auto","created_at":"2025-10-29 13:41:18","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":174630,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/f0b3167317ce2339ffa1ab9a.png"},{"id":94672673,"identity":"e30845a2-8307-432d-a7f3-bd6eec6b1216","added_by":"auto","created_at":"2025-10-29 13:40:49","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":102642,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/9c148c8aae63f5314573aff5.png"},{"id":94663751,"identity":"b94f7a63-301d-4a24-8f1a-8f55cbcb98a9","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159431,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/8697880fca9ef14a4ae4162a.png"},{"id":94672620,"identity":"02014433-023d-4daf-976a-a7f9af238326","added_by":"auto","created_at":"2025-10-29 13:40:46","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7985,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/0cefd46f7623e820aa5b7e96.png"},{"id":94672638,"identity":"b4a1e9ad-55ec-4370-8e8b-eeadd182e434","added_by":"auto","created_at":"2025-10-29 13:40:47","extension":"xml","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":98515,"visible":true,"origin":"","legend":"","description":"","filename":"03a8449c51f44934bace853f4555d2711structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/79808999a01f8133139c9014.xml"},{"id":94672512,"identity":"5fc5c409-0296-487b-9b9d-839e3b102157","added_by":"auto","created_at":"2025-10-29 13:40:40","extension":"html","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":107355,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/4a587c823692f58bc35e3873.html"},{"id":94663736,"identity":"4c8df23b-1d57-4689-8de0-b992b8431592","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":903137,"visible":true,"origin":"","legend":"\u003cp\u003eExamples of pre- processed (top) and post-processed (bottom) wing images used in training; from left to right \u003cem\u003eAustrotephritis plebeia\u003c/em\u003e, \u003cem\u003eBactrocera tryoni\u003c/em\u003e, \u003cem\u003eand Ceratitis capitata\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/9bb2d2f24fe60b2057290882.png"},{"id":94673278,"identity":"c0d2d075-f071-4512-90c9-a6d7bcc27a4f","added_by":"auto","created_at":"2025-10-29 13:41:19","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":743064,"visible":true,"origin":"","legend":"\u003cp\u003eExamples of misclassified images (left) from the test split compared to the predicted species (right). (a) Actual species \u003cem\u003eB. minuta\u003c/em\u003e, predicted species \u003cem\u003eB. passiflorae\u003c/em\u003e. (b) Actual species \u003cem\u003eB. latifrons\u003c/em\u003epredicted species \u003cem\u003eB. dorsalis\u003c/em\u003e. (c) Actual species \u003cem\u003eS. fascigera\u003c/em\u003e, predicted species \u003cem\u003eS. ruificeps\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/ed102db0a13a63d5e331cf25.png"},{"id":94663727,"identity":"783e14ae-60d3-4d29-b9f7-4566450ae5dc","added_by":"auto","created_at":"2025-10-29 12:13:11","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":55376,"visible":true,"origin":"","legend":"\u003cp\u003eConfidence distribution by prediction correctness; Incorrect (red), Correct (blue).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/5075c9f8fcb2165d00325b3a.png"},{"id":96362930,"identity":"e8ad5123-d03f-4b9c-9ba4-9e41cfb5b8d9","added_by":"auto","created_at":"2025-11-20 10:03:01","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2608591,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/a7fc6353-4556-4275-a286-cea38a24a2d6.pdf"},{"id":94663734,"identity":"d98e0e8a-ecc8-4529-a5c1-0a21df213300","added_by":"auto","created_at":"2025-10-29 12:13:12","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":1116728,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7802967/v1/967f63e52f33e4cb229417f8.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A deep learning classification pipeline for identifying economically important tephritid fruit flies","fulltext":[{"header":"Key Message","content":"\u003cul class=\"decimal_type\"\u003e\n \u003cli\u003eDeep learning boosts insect ID through image-based computer vision.\u003c/li\u003e\n \u003cli\u003eCNN accurately classifies 34 tephritid species from wing images.\u003c/li\u003e\n \u003cli\u003eYOLOv11 models achieved 0.99 detection and 0.98 classification accuracy.\u003c/li\u003e\n \u003cli\u003eResults aid biosecurity and pest management for fruit fly control.\u003c/li\u003e\n \u003cli\u003eAI shows promise to transform entomological diagnostics and surveillance.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"Introduction","content":"\u003cp\u003eTephritidae, commonly known as fruit flies, are a diverse family of Diptera that are widely distributed around the world. Several species of Tephritidae cause severe economic damage to horticulture industries in a wide range of countries through costs associated with losses to fruit and vegetable crops, the loss of market access, and the costs associated with control (Suckling et al. 2014; Dias et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Such losses can be substantial, for example, He et al. (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) estimated the economic losses in China from \u003cem\u003eZeugodacus cucurbitae\u003c/em\u003e at many hundreds of millions of dollars per year. Kibira et al. (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) noted that in eastern and southern Africa, several species of \u003cem\u003eCeratitis\u003c/em\u003e contribute to mango yield losses of between 30% to 70% depending on the locality, variety and season, and that an exotic species, \u003cem\u003eBactrocera dorsalis\u003c/em\u003e (reported as \u003cem\u003eB. invadens\u003c/em\u003e) causes damage to over 80% of mango crops. In Australia, costs associated with maintaining exclusion zones and periodic outbreaks of \u003cem\u003eB. dorsalis\u003c/em\u003e (reported as \u003cem\u003eB. papayae\u003c/em\u003e) can also be as high as hundreds of millions of dollars (Jessup et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Suckling et al. 2014).\u003c/p\u003e\u003cp\u003eConsiderable effort and expenditure are associated with different types of control and management (Dias et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; He et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), such as the use of lures (Stringer et al. 2019), area wide suppression (Lloyd et al. 2010), natural enemies (Dias et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and the sterile insect technique (Shelly and McInnis \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). However, there is also significant effort and investment in keeping countries and areas free of these pests, particularly as many species can be easily transported around the globe via human trade networks. Such investment includes quarantine and fruit destruction, surveillance networks via large scale trapping, and eradication programmes (Quilici and Donner 2012; Suckling et al. 2014; Kean et al. \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eCorrect species identification is critical, whether it is for the control and management of pests, or for keeping areas pest-free through surveillance and eradication. However, the morphological identification of tephritid species can be difficult and requires practical experience and professional skills (White \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Plant Health Australia \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Drew and Romig 2013; He et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Molecular techniques are increasingly being used to resolve species limits and distinguish species, but this is also not without problems (Dhami et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Doorenweerd et al. 2023; Starkie et al., 2024).\u003c/p\u003e\u003cp\u003eImaged-based identification offers an additional, and likely complementary, approach to morphological and molecular identification. Computer vision approaches based on machine learning and deep learning have significantly influenced a wide range of scientific disciplines, though these have only relatively recently been applied to entomology (Boer and Vos \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Marques et al. 2018; Hansen et al. 2019). Recent studies on image-based insect identification are showing that deep learning models can extract features from images and learn to differentiate species to an accuracy approaching, or exceeding, human expertise (Valan et al. \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Hoye et al. 2020, Ward and Martin \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2023\u003c/span\u003e, da Cunha et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and is being applied to a growing range of insect taxa. For example, over half of British ground beetles (Carabidae) can be identified to species, and 74% to genus using convolutional neural networks (Hansen et al. 2019). Boer and Vos (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) used over 10,000 images from AntWeb (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.antweb.org\u003c/span\u003e\u003cspan address=\"http://www.antweb.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) to classify ants at species level based on dorsal, head, and profile images, obtaining accuracy between 62\u0026ndash;92% for species and 79\u0026ndash;95% for genus, depending on different configurations of the models.\u003c/p\u003e\u003cp\u003ePrevious work on the automated identification of fruit flies is limited, however Wang et al. (2016) used custom built software (based on C# language on Microsoft\u0026rsquo;s Net Framework) to classify 74 species belonging to six genera, and despite having only a few images per species was able to obtain good identification rates depending on the morphological character, 71% (wing), 69% (thorax), 39% (abdomen). More recent work includes Tariq et al. (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) who were testing the development of identification systems for farmers and growers through mobile applications and examined the image capture of two species using smartphone and Raspberry Pi cameras in both the laboratory and fruit orchards.\u003c/p\u003e\u003cp\u003eIn this paper, we test the ability of a deep learning convolutional neural network pipeline to classify genera and species of Tephritidae using images of the wing for a range of species that are relevant around the world for pest control and management, and for border biosecurity and surveillance.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eTaxonomic species and specimens\u003c/h2\u003e\u003cp\u003eWe captured 1,380 wing images representing 34 species across 12 genera of Tephritidae, along with 1,151 images of other Diptera and 3,368 images of Hymenoptera, totalling 5,899 images. An additional 10 tephritid species were excluded from the dataset as each had fewer than 10 images available. The dataset included economically important species from the genera \u003cem\u003eBactrocera\u003c/em\u003e, \u003cem\u003eCeratitis\u003c/em\u003e, and \u003cem\u003eZeugodacus\u003c/em\u003e, native New Zealand species from \u003cem\u003eAustrotephritis\u003c/em\u003e, \u003cem\u003eSphenella\u003c/em\u003e, and \u003cem\u003eTrupanea\u003c/em\u003e, and biocontrol agents from \u003cem\u003eProcecidochares\u003c/em\u003e and \u003cem\u003eUrophora\u003c/em\u003e that have been introduced to New Zealand for invasive weed control.\u003c/p\u003e\u003cp\u003eSpecimens were primarily obtained from the New Zealand Arthropod Collection, including species from New Zealand and Pacific Island nations, using dry pinned material. Additional specimens were obtained from laboratory colonies kept at the USDA Daniel K. Inouye U.S. Pacific Basin Agricultural Research Center in Hilo, Hawaii (\u003cem\u003eB. dorsalis, Bactrocera latifrons, and Z. cucurbitae\u003c/em\u003e), the Indian Institute of Horticultural Research in Bangalore, India (\u003cem\u003eBactrocera tau\u003c/em\u003e) and South Africa (\u003cem\u003eB. dorsalis\u003c/em\u003e and \u003cem\u003eCeratitis cosyra\u003c/em\u003e). The median number of images per tephritid species was 19 (range: 10\u0026ndash;285; see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eClassification performance by species.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSpecies\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTraining Images\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eValidation Images\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTest Images\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eF1 Score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eAustrotephritis cassiniae\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eAustrotephritis plebeia\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera caliginosa\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera correcta\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera curvipennis\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera distincta\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e176\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera dorsalis\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e912\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e171\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera facialis\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e313\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera frauenfeldi\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera kirki\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e345\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera latifrons\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.73\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera melanotus\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e124\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera minuta\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera obscura\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera passiflorae\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e320\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera tau\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.89\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera tryoni\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera umbrosus\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e131\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eBactrocera xanthodes\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e211\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eCeratitis capitata\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eCeratitis cosyra\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e249\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eCoelotrypes punctilabris\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e70\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eDacus solomonensis\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eDacus vertebratus\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiptera\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e912\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e171\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eDirioxa pornia\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHymenoptera\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e912\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e171\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eProcecidochares alani\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eProcecidochares utilis\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eSphenella fascigera\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.86\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eSphenella ruficeps\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.89\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eTrupanea fenwicki\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eTrupanea longipennis\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eUrophora cardui\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eUrophora stylata\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cem\u003eZeugodacus cucurbitae\u003c/em\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e233\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eSpecimen preparation and imaging\u003c/h3\u003e\n\u003cp\u003eFor pinned specimens, wings were carefully removed from the body prior to imaging. Specimens were placed in a specimen manipulator, and a micropin or number 5 forceps was used to gently manipulate the tegula until the wing detached naturally. Wings were not forcibly pulled to avoid membrane damage. Wings were positioned on a microscope stage with a white background, imaged, and subsequently stored in gelatin capsules for preservation. For specimens from Hawaiian, Indian and South African laboratory colonies, wings were secured in position using double-sided tape during imaging. All wing images were captured using a Leica EZ4 W microscope (Leica Microsystems) with an integrated camera. The wings provided from South Africa were prepared as reported in Roets et al. (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) and Makumbe et al. (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Wings were positioned flat, eliminating the need for image stacking techniques (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eModel training and validation\u003c/h3\u003e\n\u003cp\u003eWing images were uploaded to Roboflow (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://roboflow.com/\u003c/span\u003e\u003cspan address=\"https://roboflow.com/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), an online platform for image annotation and deep learning. A project was established in Roboflow (publicly available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://universe.roboflow.com/bugider/tephritidae\u003c/span\u003e\u003cspan address=\"https://universe.roboflow.com/bugider/tephritidae\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), where polygons were drawn around each wing using the Smart Polygon tool powered by Meta AI's Segment Anything Model (SAM2). Each polygon was annotated with the corresponding taxonomic species designation.\u003c/p\u003e\u003cp\u003eImages were exported to a local workstation (HP Z2 SFF G9 with 128GB RAM and NVIDIA RTX A4000 GPU, Windows 10) for processing. Annotations were converted to bounding box format compatible with YOLO object detection models. Images underwent preprocessing including resizing to 640 \u0026times; 640 pixels, grayscale conversion (due to limited colour information in tephritid wings), and exposure correction using contrast stretching to create the object detection training dataset.\u003c/p\u003e\u003cp\u003eWe used YOLOv11 models from the Ultralytics Python library (Jocher et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). We trained a two-stage model, first with a detection stage (finding the wing in the image) and then a classification stage (identification to species). A single object detection model was trained treating all wings as a single class. Multiple classification models were trained with varying dropout rates (0.2\u0026ndash;0.5) and learning rates (0.0001\u0026ndash;0.01) to optimize performance. Complete training parameters are detailed in Supplementary Table\u0026nbsp;1. Both datasets (detection and classification) were partitioned into training (85%), validation (15%), and test (5%) splits using stratified sampling to maintain proportional class representation across splits. For the classification dataset (i.e. stage two cropped wings generated from the detection stage), images were further processed by augmenting through random flipping, blurring, and noise addition. The two 'other' classes (Hymenoptera and Diptera) were randomly subsampled to match the size of the largest tephritid class, preventing training bias. The final classification dataset comprised 7,800 images.\u003c/p\u003e\u003cp\u003eModel performance was evaluated on the test splits in each image set (unseen during training or validation) using mean Average Precision (mAP) for object detection and accuracy, precision, recall, and F1 score for species classification. We also assessed performance by comparing the confidence scores of correct versus incorrect predictions on the test set.\u003c/p\u003e"},{"header":"Results and Discussion","content":"\u003cp\u003eBoth the detection and classification models converged well after 30 training epochs. Training and validation loss curves were similar for both models (Supplementary Fig.\u0026nbsp;1).\u003c/p\u003e\n\u003ch3\u003eObject detection performance\u003c/h3\u003e\n\u003cp\u003eThe model (YOLOv11n) demonstrated excellent performance in detecting wing objects within images, achieving precision and recall of 1.00 for both metrics on the test dataset. The mean Average Precision at IoU threshold 0.5 (mAP50) was 1.00 and at mAP50-95 was 0.99. These results indicate that the model was highly effective at locating and delineating wing boundaries across all images, successfully distinguishing wing objects from the background. While these results may appear unrealistically high, it should be noted that wings were always displayed on a plain background, there was a single wing per image, and all classes were lumped for the object detection task, so precision and recall were expected to be very high. Further, we would expect that in any deployment of this model (i.e. for identification queries) the images would be captured in very similar conditions to the training images to ensure similarly high detection rates (see Methods). Images based from specimens photographed in field situations or of the whole body (wings not removed) are not expected to perform as well.\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eSpecies classification performance\u003c/h2\u003e\u003cp\u003eThe classification model (YOLOv11n-cls) showed strong overall performance across the 34 tephritid species and two other taxonomic groups. Overall (across classes) classification accuracy was 0.98, with precision of 0.98, recall of 0.97, and F1 score of 0.97, demonstrating robust species-level identification.\u003c/p\u003e\u003cp\u003eClassification performance varied considerably among species, revealing patterns related to dataset composition and morphological distinctiveness (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Twenty-six species (76%) achieved perfect classification metrics (accuracy, precision, recall, and F1 scores of 1.00), including economically important pest species such as \u003cem\u003eB. dorsalis\u003c/em\u003e, \u003cem\u003eB. tryoni\u003c/em\u003e, \u003cem\u003eC. capitata\u003c/em\u003e, and \u003cem\u003eZ. cucurbitae\u003c/em\u003e. Several species showed slight reductions in precision while maintaining perfect recall and accuracy. \u003cem\u003eBactrocera passiflorae\u003c/em\u003e achieved 0.91 precision, \u003cem\u003eB. facialis\u003c/em\u003e recorded 0.95 precision, and \u003cem\u003eSphenella ruficeps\u003c/em\u003e showed 0.80 precision. These lower precision values suggest occasional misclassification of other species (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Supplementary Fig.\u0026nbsp;2).\u003c/p\u003e\u003cp\u003eThe most challenging classifications involved species with smaller training datasets and that were visually like other species in the same genus. \u003cem\u003eBactrocera minuta\u003c/em\u003e (83 training images) achieved 67% accuracy with precision of 1.00 but recall of 0.67, indicating that while wings predicted as \u003cem\u003eB. minuta\u003c/em\u003e were accurately classified, some individuals within this class were misassigned to the visually similar \u003cem\u003eB. passiflorae\u003c/em\u003e (e.g. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Similarly, \u003cem\u003eB. latifrons\u003c/em\u003e (96 training images) showed 67% accuracy with 0.80 precision and 0.67 recall (misclassified as \u003cem\u003eB. dorsalis\u003c/em\u003e, e.g. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb), \u003cem\u003eB. tau\u003c/em\u003e (57 training images) achieved 80% accuracy with precision of 1.0 and 0.8 recall (misclassified as \u003cem\u003eB. latifrons\u003c/em\u003e), and \u003cem\u003eSphenella fascigera\u003c/em\u003e (44 training images) achieved 75% accuracy with precision of 1.0 and 0.75 recall (misclassified as \u003cem\u003eS. ruificeps\u003c/em\u003e, e.g. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec). \u003cem\u003eBactrocera kirki\u003c/em\u003e, despite having a larger training set (345 images), showed slightly reduced performance (96% accuracy, 1.00 precision, 0.96 recall, misclassified as \u003cem\u003eB. fascialis\u003c/em\u003e), suggesting morphological similarity to closely related species rather than sample size limitations. All misclassified images were assigned to other species within the same genus as the source species. Wing morphology is also much more conserved within genera than across genera.\u003c/p\u003e\u003cp\u003eThe high performance across diverse genera, including functionally distinct groups such as agricultural pests (e.g. \u003cem\u003eBactrocera\u003c/em\u003e, \u003cem\u003eCeratitis\u003c/em\u003e), biocontrol agents (\u003cem\u003eProcecidochares\u003c/em\u003e, \u003cem\u003eUrophora\u003c/em\u003e) and native New Zealand species (\u003cem\u003eAustrotephritis\u003c/em\u003e, \u003cem\u003eTrupanea\u003c/em\u003e), demonstrates the classification model's ability to generalise across varied wing morphologies to accurately identify species of economic importance. Further, there was generally good separation in confidence scores between correct and incorrect predictions, with correct predictions skewed towards 1.0 and incorrect predictions typically below 0.8 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this study we demonstrated the successful application of deep learning methods for automated species-level identification of Tephritidae, achieving excellent accuracy (98%) in distinguishing among 34 fruit fly species using wing morphology. The two-stage YOLOv11 pipeline showed perfect object detection performance and robust classification across economically critical pest species including \u003cem\u003eB. dorsalis\u003c/em\u003e, \u003cem\u003eB. tryoni\u003c/em\u003e, \u003cem\u003eC. capitata\u003c/em\u003e, and \u003cem\u003eZ. cucurbitae\u003c/em\u003e, which achieved near-perfect identification rates. Our results represent a significant advancement for biosecurity and pest management applications, where rapid and accurate species identification is essential for effective quarantine decisions, surveillance programs, and targeted control strategies. The classification model's ability to reliably distinguish between pest species, beneficial biocontrol agents, and native taxa addresses a critical bottleneck in entomological diagnostics, particularly in resource-limited settings where taxonomic expertise may be unavailable. By providing an accessible, automated identification tool that maintains high accuracy across diverse geographic specimens, this technology has the potential to transform frontline biosecurity operations. The further development of desktop or mobile applications will also enable real-time pest detection at ports of entry and support management decisions that protect agricultural systems and native ecosystems from invasive species.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eCompeting Interests\u003c/h2\u003e\u003cp\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003eThis work was supported by SSIF Infrastructure funding for Nationally Significant Collections and Databases at Manaaki Whenua-Landcare Research; the B3 (Better Border Biosecurity) project D22.15 on \u0026lsquo;Using images and deep learning for the identification of high-risk insect species\u0026rsquo;; and the University of Auckland Summer Research program for Tomas Blokker, Angie Zhu, and Hope Ryu.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eDW and AH conceived and designed research. DW, AH, CW, VS, and DPS provided images. DB, KW and BM provided computer science support. DW and AH wrote the manuscript. All authors read and approved the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003ePart of the images was made possible thanks to the Cooperative Agreements (23-8130-1011-IA, 25-8130-1011-CA) from the United States Department of Agriculture\u0026rsquo;s Animal and Plant Health Inspection Service (APHIS) awarded to DPS. It may not necessarily express APHIS\u0026rsquo; views. DPS thanks Abraham Moreno Mej\u0026iacute;a and Kamala Jayanthi. Thanks to Tomas Blokker, Angie Zhu, and Hope Ryu at the University of Auckland for taking images.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eSpecimen records are accessible through GBIF (https://www.gbif.org) or Symbiota's Ecdysis portal (https://ecdysis.org). Wing images are available via Roboflow (https://universe.roboflow.com/bugider/tephritidae), and Python scripts are provided on GitHub ( [https://github.com/aharmer/tephritID](https:/github.com/aharmer/tephritID) https://github.com/aharmer/tephritID).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003e\u0026Auml;rje J, Melvad C, Jeppesen MR, Madsen SA, Raitoharju J, Rasmussen MS, Iosifidis A, Tirronen V, Gabbouj M, Meissner K, H\u0026oslash;ye TT (2020). Automatic image-based identification and biomass estimation of invertebrates. Methods in Ecology and Evolution 11:922\u0026ndash;931.\u003c/li\u003e\n\u003cli\u003eBjerge K, Nielsen JB, Sepstrup MV, Helsing-Nielsen F, H\u0026oslash;ye TT (2021). An automated light trap to monitor moths (Lepidoptera) using computer vision-based tracking and deep learning. Sensors 21:343.\u003c/li\u003e\n\u003cli\u003eBoer MJ, Vos RA (2018). Taxonomic classification of ants (Formicidae) from images using deep learning. bioRxiv 407452.\u003c/li\u003e\n\u003cli\u003eda Cunha VAG, Pullock DA, Ali M, Neto AdOC, Ampatzidis Y, Weldon CW, Kruger K, Manrakhan A, Qureshi, J (2024). Psyllid Detector: A web-based application to automate insect detection utilizing image processing and deep learning. Applied Engineering in Agriculture 40:427\u0026ndash;438.\u003c/li\u003e\n\u003cli\u003eDhami MK, Gunawardana DN, Voice D, Kumarasinghe L (2016). A real-time PCR toolbox for accurate identification of invasive fruit fly species. Journal of Applied Entomology 140:536\u0026ndash;552. https://doi.org/10.1111/jen.12286\u003c/li\u003e\n\u003cli\u003eDias NP, Montoya P, Nava DE (2022). A 30-year systematic review reveals success in tephritid fruit fly biological control research. Entomologia Experimentalis et Applicata 170:370\u0026ndash;384. https://doi.org/10.1111/eea.13157\u003c/li\u003e\n\u003cli\u003eDias NP, Zotti MJ, Montoya P, Carvalho IR, Nava DE (2018). Crop protection, fruit fly management research: a systematic review of monitoring and control tactics in the world. Crop Protection 112:187\u0026ndash;200.\u003c/li\u003e\n\u003cli\u003eDoorenweerd C, Leblanc L, Norrbom AL, Jose MS, Rubinoff D (2018). A global checklist of the 932 fruit fly species in the tribe Dacini (Diptera, Tephritidae). Zookeys 730:19\u0026ndash;56. \u003c/li\u003e\n\u003cli\u003eDrew RAI, Romig M (2013). Tropical Fruit Flies of South-East Asia (Tephritidae: Dacinae): Indomalaya to North-West Australasia. CABI, Wallingford. https://doi.org/10.1079/9781780640358.0000\u003c/li\u003e\n\u003cli\u003eGreeff M, Caspers M, Kalkman V, Willemse L, Sunderland BD, B\u0026aacute;nki O, Hogeweg L (2022). Sharing taxonomic expertise between natural history collections using image recognition. Research Ideas and Outcomes 8:e79187.\u003c/li\u003e\n\u003cli\u003eHansen OL, Svenning JC, Olsen K, Dupont S, Garner BH, Iosifidis A, Price BW, H\u0026oslash;ye TT (2020). Species‐level image classification with convolutional neural network enables insect identification from habitus images. Ecology and Evolution 10:737\u0026ndash;747.\u003c/li\u003e\n\u003cli\u003eHe Y, Xu Y, Chen X (2023). Biology, ecology and management of tephritid fruit flies in China: a review. Insects 14:196. https://doi.org/10.3390/insects14020196\u003c/li\u003e\n\u003cli\u003eH\u0026oslash;ye TT, \u0026Auml;rje J, Bjerge K, Hansen OLP, Iosifidis A, Leese F, Mann HMR, Meissner K, Melvad C, Raitoharju J (2021). Deep learning and computer vision will transform entomology. Proceedings of the National Academy of Sciences 118(2):e2002545117.\u003c/li\u003e\n\u003cli\u003eJessup AJ, Dominiak B, Woods B, De Lima CPF, Tomkins A, Smallridge CJ (2007). Area-Wide Management of Fruit Flies in Australia. In: Vreysen, M.J.B., Robinson, A.S., Hendrichs, J. (eds) Area-Wide Control of Insect Pests. Springer, Dordrecht. https://doi.org/10.1007/978-1-4020-6059-5_63\u003c/li\u003e\n\u003cli\u003eJocher G, Qiu J, Chaurasia, A (2023). Ultralytics YOLO (Version 8.0.0) [Computer software]. https://github.com/ultralytics/ultralytics\u003c/li\u003e\n\u003cli\u003eKean JM, Manoukis NC, Dominiak BC (2024). Review of surveillance systems for tephritid fruit fly threats in Australia, New Zealand, and the United States. Journal of Economic Entomology 117:8\u0026ndash;23. https://doi.org/10.1093/jee/toad228 \u003c/li\u003e\n\u003cli\u003eKibira M, Affognon H, Njehia B, Muriithi B, Mohamed S, Ekesi S (2010). Economic evaluation of integrated management of fruit fly in mango production in Embu County, Kenya. African Journal of Agricultural and Resource Economics 10:343\u0026ndash;353.\u003c/li\u003e\n\u003cli\u003eMakumbe LDM, Moropa TP, Manrakhan A, Weldon CW (2020). Effect of sex, age and morphological traits on tethered flight of \u003cem\u003eBactrocera dorsalis\u003c/em\u003e (Hendel) (Diptera: Tephritidae) at different temperatures. Physiological Entomology 45:110-119.\u003c/li\u003e\n\u003cli\u003ePlant Health Australia (2011). The Australian Handbook for the Identification of Fruit Flies. Version 1.0. Plant Health Australia. Canberra, ACT.\u003c/li\u003e\n\u003cli\u003eRoets PD, Bosua H, Ruth Archer C, Weldon CW (2018). Life-history and demographic traits of the marula fruit fly \u003cem\u003eCeratitis cosyra\u003c/em\u003e: potential consequences of host specialization. Physiological Entomology 43:259-267.\u003c/li\u003e\n\u003cli\u003eShelly T, McInnis D (2016). Sterile insect technique and control of tephritid fruit flies: do species with complex courtship require higher overflooding ratios? Annals of the Entomological Society of America 109:1\u0026ndash;11. https://doi.org/10.1093/aesa/sav101\u003c/li\u003e\n\u003cli\u003eTariq S, Hakim A, Siddiqi AA, Owais M (2022). An image dataset of fruitfly species (Bactrocera zonata and Bactrocera dorsalis) and automated species classification through object detection. Data Brief 43:108366. https://doi.org/10.1016/j.dib.2022.108366\u003c/li\u003e\n\u003cli\u003eWang J, Chen X, Hou X, Zhou L, Zhua C, Jia L (2017). Construction, implementation and testing of an image identification system using computer vision methods for fruit flies with economic importance (Diptera: Tephritidae). Pest Management Science 73:1511\u0026ndash;1528. https://doi.org/10.1002/ps.4487\u003c/li\u003e\n\u003cli\u003eWard DF, Martin B (2023). Trialling a convolution neural network for the identification of Braconidae in New Zealand. Journal of Hymenoptera Research 95:95\u0026ndash;101.\u003c/li\u003e\n\u003cli\u003eWhite IM (2006). Taxonomy of the Dacina (Diptera: Tephritidae) of Africa and the Middle East. African Entomology, Memoir No. 2, 156 pp.\u003c/li\u003e\n\u003cli\u003eValan M, Makonyi K, Maki A, Vondr\u0026aacute;cek D, Ronquist F (2019). Automated taxonomic identification of insects with expert-level accuracy using effective feature transfer from convolutional networks. Systematic Biology 68:876\u0026ndash;895. \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"Biosecurity, Computer vision, Insects, Pest management, Surveillance, Tephritidae","lastPublishedDoi":"10.21203/rs.3.rs-7802967/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7802967/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eComputer vision approaches utilising deep learning offer significant potential benefits for entomological applications, particularly for image-based taxonomic identification. Fruit flies (Tephritidae) represent economically damaging pests where species-level identification is critical for effective pest control, management, surveillance, and eradication programs. We assessed the capability of a deep learning convolutional neural network (CNN) pipeline to identify tephritid species from wing images, which serve as key diagnostic features. Our dataset comprised 1380 tephritid wing images spanning 34 tephritid species and 12 genera, with additional images from two 'other' classes (other Diptera and Hymenoptera). We employed a two-stage approach: (1) object detection using an Ultralytics YOLOv11n model to detect wing objects in images, treating all wings as a single class, followed by (2) species classification using an Ultralytics YOLOv11-cls model applied to cropped and augmented wing images generated from the object detection stage. The models demonstrated high accuracy in both wing detection (mAP50-95 value of 0.99 on a novel test set) and species classification (overall accuracy of 0.98 on a novel test set). Class-wise accuracy for different species varied (0.67-1) but showed general correlation with the number of original images available per class (10\u0026ndash;285). Our results provide a potentially valuable tool for detecting pest tephritid species in biosecurity contexts. While deep learning technology remains in early development stages for entomological applications, such approaches hold promise to transform diagnostic and surveillance capabilities for biosecurity and pest management.\u003c/p\u003e","manuscriptTitle":"A deep learning classification pipeline for identifying economically important tephritid fruit flies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-29 12:13:07","doi":"10.21203/rs.3.rs-7802967/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":"15c3fb64-ee93-416b-b5f0-ac83cc7cd040","owner":[],"postedDate":"October 29th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-11-14T15:38:46+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-29 12:13:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7802967","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7802967","identity":"rs-7802967","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}