FishProx: Proximate conspecific interaction of Atlantic salmon (Salmo salar) for behavioural analysis through instance segmentation masks

FishProx: Proximate conspecific interaction of Atlantic salmon (Salmo salar) for behavioural analysis through instance segmentation masks | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (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];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results FishProx: Proximate conspecific interaction of Atlantic salmon ( Salmo salar ) for behavioural analysis through instance segmentation masks View ORCID Profile I-Hao Chen , View ORCID Profile Nabil Belbachir , View ORCID Profile Lars Ebbesson , Antonella Zanna Munthe-Kaas , View ORCID Profile Lars Erik Solberg , View ORCID Profile Gaute Alexander Nedberg Helberg , David Izquierdo-Gomez , Santhosh K. Kumaran , View ORCID Profile Jelena Kolarevic , View ORCID Profile Chris Noble doi: https://doi.org/10.1101/2025.07.07.659286 I-Hao Chen 1 Fish Biology & Aquaculture Group, NORCE Norwegian Research Centre AS , Bergen, Norway 2 Department of Mathematics, University of Bergen Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for I-Hao Chen For correspondence: chen{at}norceresearch.no Nabil Belbachir 3 DARWIN Group, NORCE Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Nabil Belbachir Lars Ebbesson 1 Fish Biology & Aquaculture Group, NORCE Norwegian Research Centre AS , Bergen, Norway Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Lars Ebbesson Antonella Zanna Munthe-Kaas 2 Department of Mathematics, University of Bergen Find this author on Google Scholar Find this author on PubMed Search for this author on this site Lars Erik Solberg 4 Nofima AS Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Lars Erik Solberg Gaute Alexander Nedberg Helberg 5 The Norwegian College of Fishery Science, UiT The Arctic University of Norway Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Gaute Alexander Nedberg Helberg David Izquierdo-Gomez 4 Nofima AS Find this author on Google Scholar Find this author on PubMed Search for this author on this site Santhosh K. Kumaran 4 Nofima AS Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jelena Kolarevic 4 Nofima AS 5 The Norwegian College of Fishery Science, UiT The Arctic University of Norway Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jelena Kolarevic Chris Noble 4 Nofima AS Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Chris Noble Abstract Full Text Info/History Metrics Preview PDF Abstract In modern fish production and experimental facilities, camera installations are increasingly common for monitoring animal welfare. This opens up a new alley of autonomous tracking systems levering streamed data. However, the substantial amount of data necessitates condensation for human operators. In this paper, we present the out-of-the-box pipeline called FishProx that starts with the imagery of fish in tanks and ends in the calculation of welfare impacting factors through anomaly scores. The utilisation of the Segment Anything Model (SAM) is the basis for the identification of feeding behaviour and also (ab)normal behaviour through analysing snapshots of the fish tanks to infer metrics such as fish cohesion and cluster alignment, alongside the detection count. Our method employs automatic and robust algorithms for camera systems, making additional labelled data unnecessary, thereby saving energy and costs. Furthermore, automatic fish observation coupled with integrated analytical capabilities can lead to a sensible autonomous decision-making process. Lastly, by providing objective metrics, our approach might guide the enabling of automatic quantification of fish states, making comparative studies possible. 1. Introduction “Fish that are close to each other, should swim together”. This is the idea leading to the pipeline FishProx, which implements a microservice-oriented architecture to utilise the Segment Anything Model (SAM) [ 9 ] to facilitate downstream tasks like local fish behaviour analysis. Download figure Open in new tab Figure 1. Schematics of FishProx: The segment anything model (SAM) [ 9 ] is used on extracted frames to prompt segmentation masks that are filtered for fish segmentations. Afterwards, the segmentation centers and their 180 ° -agnostic orientation are calculated. To obtain anomaly scores, segementation centers are clustered with K-means (K=16) which gives rise to the fish cohesion score (intra-cluster) and the cluster alignment score (inter-cluster) using the in-and outliers of the orientation angles. Integrating these scores with the segmentation detection count enables the quantification of some fish behaviour in RAS tanks. 1.1 Related Work In the past few years, generative AI is hitting our mainstream society around the world [ 4 , 13 , 22 ]. Natural language processing (NLP) [ 5 ], generative pre-trained transformers (GPT) [ 19 , 26 ] and large language models (LLM)[ 18 ] have gained in maturity over the past years to show impressive capabilities in the everyday life as tools ranging from education [ 3 , 15 , 16 ] to content creation [ 24 ]. At the same time, generative models for animal behaviour are not widespread. The labelling and tracking of animals [ 14 ] for various purposes have been explored, though not for fish. Work on zero-shot detection for species recognition can be found [ 21 ], but to the best of our knowledge, there exists no integrated pipeline or model that generates meaningful metrics for fish populations from video data. Work on fish segmentation itself, on the other hand, has been studied using various architectures [ 7 , 10 , 23 , 25 ], and with the rise of the Segment Anything Model (SAM) [ 9 ], the way segmentation can be obtained was also revolutionised for the aquatic sector. 1.2. Recirculating aquaculture system (RAS) Recirculating aquaculture systems (RAS) have been under heavy development in the past two decades, which lead to increased size and fish capacity and optimized resource usage [ 6 ]. Using RAS to farm fish has advantages in the environment-friendly and sustainable area [ 11 ]. It is water efficient and highly productive and is often protected from environmental impacts such as pollution and parasitic infections [ 2 ]. Current trends suggest an increase of RAS facilities around the world inline with an increase in production of Atlantic salmon Salmo salar ), one of the major farmed fish species [ 1 ]. 1.3. Experiment Details The RAS research facility used in this experiment was located in Sunndalsøra, Norway. Each of the six water tanks had a volume of 3.3 m 3 and was shaped like an octagon. Additionally, a top camera, a IR Dome (Imenco AS) was installed above each tank with a video resolution of 1080p @ 30fps. Another camera was placed underwater, and a hydrophone (the long rod through the pictures in Figure 2 ) was installed for each tank. A trial experiment was ongoing for several months in 2023 (May until August), which is not the focus of this paper. A total of 1000 Atlantic salmon were in each tank with a weight of 40 grams ± 10 grams at the starting date of 31.03.2023. The data available for this study was recorded at tank 210 and consists of around 8 hours of video material from 12.05.2023. Download figure Open in new tab Figure 2. In sequential order: Original image of the fish RAS tank 210, the tank with fish segmentation based on SAM [ 9 ], then orientation angles are displayed for each fish segmentation, and lastly, the classification that leads to anomaly scores (see Section 2.4.1 ). 2. The FishProx pipeline The FishProx pipeline consists of multiple Python code modules that were put in docker images [ 17 ] and orchestrated by a dockerfile. The running services were communicating through the file-system which allowed for manual inspection. Chained together, they formed a robust video processing pipeline which could be used OS independently. We present the modules in order: 2.1. Video Extraction While this service was running, new videos were just dropped into its overseen folder, triggering the processing into frames in a parallel manner. 2.2. Frame Analysis Frames added to the input folder of this service triggered SAM [ 9 ] (checkpoint sam vit l 0b3195) to prompt segmentation masks for the frames. This was programmed concurrently with shared GPU such that we achieve full utilisation of the GPU (Nvidia GeForce RTX4090) as this step was usually the bottleneck. The segmentation result for each frame from SAM was saved into the file system as a .json file. 2.3. Mask Processing 2.3.1 Mask Selection From all the segmentations of the fish tank, only fish segmentations were of interest. A filtering technique based on Kernel Density Estimation (KDE) [ 20 ] was applied to the pixel size of the segmentations in conjunction with creating KDE “groups” through thresholding with local extrema. Here we are leveraging knowledge about the fish size of the same rearing process which have similar sizes and therefore being grouped into the same KDE “group”. Additionally, the convex hull of the tank segmentation was calculated to account for the other objects occluding the water surface and then used as a condition for the fish masks. The center of each fish segmentation was calculated by taking the center of mass of the segmentation. Additionally, each fish segmentation was tagged with the water tank section it appeared in. The tank sections ranged from 1 to 8 and were evenly anti-clockwise distributed around the centrally located water outlet in the RAS tank. Section One was marked by the 45 ° slice by taking the right horizontal from the water outlet and rotating it anti-clockwise. Outputs of this service were the center points of each detected fish individual and the finished segmented masks for each frame which gave rise to the first metric: The number of detected fish in the frame, alias the detection count score. 2.3.2 Calculation of Orientation Angles PCA [ 8 ] was used to obtain the main axis of fish for orientation. The biggest Eigenvector ( x, y ) was used to calculate the orientation angle: The resulting angles were in the interval of [0 ° ,360 ° ). We did not differentiate between forward or backward movements, since we considered both swimming with and against the water currents valid swimming directions. Therefore, we applied the surjection mod 180 : → [0, 360) [0, 180) with x → x mod 180 to overlap the angles are 180 ° apart. This resulted in a 180 ° agnostic orientation and the angle value domain was modelled by the interval [0 ° , 180 ° ). A fish that had the orientation of a vertical line (facing upwards or downwards) thereby had an angle of 0 ° , and a fish with horizontal alignment had an angle 90 ° . 2.4. Checking Proximate Swimming Behaviour We clustered the fish with K-means [ 12 ] with K = 16 and for each cluster k ∈ {1 , …, 16}, the family of N k orientation angles were taken to calculate the circular MAD (Median Absolute Distance). To preserve the 180 ° periodicity, we normalised the orientation angles to radians with the scaling factor 2 π/ 180 to calculate their position on the unit circle with sine and cosine. Then, the median of the sines and cosines was calculated. We note that we made use of the notation . For each cluster group k we calculated: and then re-transforming the median values back into an angle Lastly, we add 180 if the c median ( k ) is negative to push all values into the positives. For each cluster element in cluster k , the median absolute distance was calculated component-wise using m sin ( k ) and m cos ( k ) as medians and calculating in the distance after transforming the angle of the cluster element into a point on the unit circle: Fish whose orientation angles fell outside of the interval c normal ( k ) := [ c median ( k ) − c MAD ( k ), c median ( k ) + c MAD ( k )] were marked as anomalous (red) for swimming different from their close peers. Otherwise, fish that swim uniformly by the same criteria were marked normal (green) (see Figure 2 ). 2.4.1 Anomaly Scores The fish cohesion score was calculated by dividing the number of normal swimming fish by the total number of fish segmentations. The cluster alignment was calculated by dividing the number of K-means clusters with zero anomalous segmentation centers by the number of all clusters. 3. Results The results were aggregated in one-minute intervals. 3.1. Feeding Behaviour Automatic feeding of fish started at 9:00 and ended at 11:00 on the 12th of May 2023, as marked in green in Figure 3 and 4 . There was only a slight increase in the fish cohesion at the start of the feeding time, but cluster alignment spiked up at the feeding start at 9:00 and persisted for half an hour before normalising itself after an hour. The number of detections dropped sharply to around half of their previous level in Figure 4 but also recovered within the next hour. Especially the sections five to eight which lied close to the feeder position display a heavy drop in counts. Note that the metrics of fish cohesion and cluster alignment are inverse to the detection count. Download figure Open in new tab Figure 3. Fish cohesion and cluster alignment as anomaly scores. The green horizontal line indicates the feeding period, and the vertical orange line the tank cleaning event at 11:17. Download figure Open in new tab Figure 4. Detection counts by tank section. The green horizontal line indicates the feeding period, and the vertical orange line the tank cleaning event at 11:17. 3.2. Disturbance Event One special datapoint of interest was 11:17. A disturbance to the fish group was captured on video when the fish population rapidly evacuated one area of the tank due to the in tank cleaning operations carried out by a member of staff. This behavioural response can be indicative of exposure to a stressor. This single event was also visible in Figure 3 , being the most prominent spike in both fish cohesion and cluster alignment. 4. Discussion Fish segmentation is a difficult task, but with stable conditions as in RAS, we can analyse video streams efficiently. It also allows to use of strong reflection and unresting water as a source of information, leading to the metric of counting the number of fish detections. We were able to show that both feeding and tank cleaning events could be measured with the presented metrics. We also showed that the disturbance due to tank cleaning leads to a higher spike, meaning a stronger reaction of fish in the tank was captured accordingly by the pipeline. The inverse relationship between detection counts and the anomaly scores is expected, since when the number of detections in total drops, and we assume an even distribution of cluster centers by K-means, then each cluster will have fewer fish individuals, making it more likely to be in the orientation angle interval in Section 2.4 . The main drawback of this pipeline is the horizontal scalability and reliance on stable videos. Bigger tank setups may require different camera positioning or it may not be feasible to have the whole tank in scope. 5. Conclusion In this study, we presented the deployable pipeline FishProx for fish state observation. We hope to push further the digitisation of the aquaculture industry with the presented readable metrics by providing quantification of events such as feeding and disturbances. Insights on anomalous behaviour in fish held in RAS can be useful to a more holistic view on fish welfare which is one of the steps towards an automated welfare assessment system. Further work would encompass the implementation of real-time sensors with pipelines like FishProx to make datastreams easily accessible and interpretable for facility personnel. 6. Funding This work was funded by the RAS4.0 project through Research Council of Norway (RCN) under the project number 320717 and the project grant with number 323300. The authors would like to thank Britt Kristin Megård Reiten for the technical expertise and assistance during the study. Funder Information Declared The Research Council of Norway, https://ror.org/00epmv149 , 320717 , 323300 References [1]. ↵ The State of World Fisheries and Aquaculture 2020 . FAO , 2020 . 2 [2]. ↵ Nesar Ahmed and Giovanni M. Turchini . Recirculating aquaculture systems (RAS): Environmental solution and climate change adaptation . Journal of Cleaner Production , 297 : 126604 , 2021 . 2 [3]. ↵ Mohammad Awad AlAfnan , Samira Dishari , Marina Jovic , and Koba Lomidze . ChatGPT as an Educational Tool: Opportunities, Challenges, and Recommendations for Communication, Business Writing, and Composition Courses . Journal of Artificial Intelligence and Technology , 3 ( 2 ): 60 – 68 , 2023 . Number: 2. 2 OpenUrl [4]. ↵ A. Bozkurt , J. Xiao , S. Lambert , A. Pazurek , H. Crompton , S. Koseoglu , R. Farrow , M. Bond , C. Nerantzi , S. Honey-church , M. Bali , J. Dron , K. Mir , B. Stewart , E. Costello , J. Mason , C. M. Stracke , E. Romero-Hall , A. Koutropoulos , C. M. Toquero , L. Singh , A. Tlili , K. Lee , M. Nichols , E. Ossiannilsson , M. Brown , V. Irvine , J. E. Raffaghelli , G. Santos-Hermosa , O. Farrell , T. Adam , Y. L. Thong , S. SaniBozkurt , R. C. Sharma , S. Hrastinski , and P. Jandrić . Speculative futures on ChatGPT and generative artificial intelligence (AI): A collective reflection from the educational landscape . Asian Journal of Distance Education , 18 ( 1 ): 53 – 130 , 2023 . 2 OpenUrl [5]. ↵ K. R. Chowdhary . Natural Language Processing . In Fundamentals of Artificial Intelligence , pages 603 – 649 . Springer India, New Delhi , 2020 . 2 [6]. ↵ Carlos A. Espinal and Daniel Matulić . Recirculating Aquaculture Technologies . In Aquaponics Food Production Systems: Combined Aquaculture and Hydroponic Production Technologies for the Future , pages 35 – 76 . Springer International Publishing , Cham , 2019 . 2 [7]. ↵ Yanling Han , Bowen Zheng , Xianghong Kong , Junjie Huang , Xiaotong Wang , Tianhong Ding , and Jiaqi Chen . Underwater Fish Segmentation Algorithm Based on Improved PSPNet Network . Sensors , 23 ( 19 ): 8072 , 2023 . Number: 19 Publisher: Multidisciplinary Digital Publishing Institute . 2 OpenUrl PubMed [8]. ↵ Ian Jolliffe . Principal Component Analysis . In International Encyclopedia of Statistical Science , pages 1094 – 1096 . Springer , Berlin, Heidelberg , 2011 . 3 [9]. ↵ Alexander Kirillov , Eric Mintun , Nikhila Ravi , Hanzi Mao , Chloe Rolland , Laura Gustafson , Tete Xiao , Spencer White-head , Alexander C. Berg , Wan-Yen Lo , Piotr Dollár , and Ross Girshick . Segment Anything , 2023 . arXiv: 2304.02643 [cs]. 1, 2, 3 [10]. ↵ Thanh Viet Le , Van Yem Vu , Van Truong Pham , and Thi-Thao Tran . A Fully Convolutional Network with Waterfall Atrous Spatial Pooling and Localized Active Contour Loss for Fish Segmentation . EAI Endorsed Transactions on Industrial Networks and Intelligent Systems , 10 ( 1 ): e4 – e4 , 2023 . Number: 1. 2 OpenUrl [11]. ↵ Hao Li , Zhengguo Cui , Hongwu Cui , Ying Bai , Zhendong Yin , and Keming Qu . A review of influencing factors on a recirculating aquaculture system: Environmental conditions, feeding strategies, and disinfection methods . Journal of the World Aquaculture Society , 54 ( 3 ): 566 – 602 , 2023 . eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jwas.12976 . 2 OpenUrl [12]. ↵ J. MacQueen . Some methods for classification and analysis of multivariate observations . In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Statistics , pages 281 – 298 . University of California Press , 1967 . 3 [13]. ↵ Nishith Reddy Mannuru , Sakib Shahriar , Zoë A Teel , Ting Wang , Brady D Lund , Solomon Tijani , Chalermchai Oak Pohboon , Daniel Agbaji , Joy Alhassan , JaKLyn Galley , Raana Kousari , Lydia Ogbadu-Oladapo , Shubham Kumar Saurav , Aishwarya Srivastava , Sai Priya Tummuru , Sravya Uppala , and Praveenkumar Vaidya . Artificial intelligence in developing countries: The impact of generative artificial intelligence (AI) technologies for development . Information Development , page 02666669231200628 , 2023 . Publisher: SAGE Publications Ltd . 2 [14]. ↵ Alexander Mathis , Pranav Mamidanna , Kevin M. Cury , Taiga Abe , Venkatesh N. Murthy , Mackenzie Weygandt Mathis , and Matthias Bethge . DeepLabCut: markerless pose estimation of user-defined body parts with deep learning . Nature Neuroscience , 21 ( 9 ): 1281 – 1289 , 2018 . Publisher: Nature Publishing Group . 2 OpenUrl CrossRef PubMed [15]. ↵ Bahar Memarian and Tenzin Doleck . ChatGPT in education: Methods, potentials, and limitations . Computers in Human Behavior: Artificial Humans , 1 ( 2 ): 100022 , 2023 . 2 OpenUrl CrossRef [16]. ↵ Marta Montenegro-Rueda , José Fernández-Cerero , José María Fernández-Batanero , and Eloy López-Meneses . Impact of the Implementation of ChatGPT in Education: A Systematic Review . Computers , 12 ( 8 ): 153 , 2023 . Number: 8 Publisher: Multidisciplinary Digital Publishing Institute . 2 OpenUrl [17]. ↵ Amit M Potdar , Narayan Dg , Shivaraj Kengond , and Mohammed Moin Mulla . Performance Evaluation of Docker Container and Virtual Machine . Procedia Computer Science , 171 : 1419 – 1428 , 2020 . 2 OpenUrl [18]. ↵ Mohaimenul Azam Khan Raiaan , Md Saddam Hossain Mukta , Kaniz Fatema , Nur Mohammad Fahad , Sadman Sakib , Most. Marufatul Jannat Mim , Jubaer Ahmad , Mohammed Eunus Ali , and Sami Azam . A Review on Large Language Models: Architectures, Applications, Taxonomies , Open Issues and Challenges. preprint, Preprints , 2023 . 2 [19]. ↵ Partha Pratim Ray . ChatGPT: A comprehensive review on background, applications, key challenges, bias, ethics, limitations and future scope . Internet of Things and Cyber-Physical Systems , 3 : 121 – 154 , 2023 . 2 OpenUrl CrossRef [20]. ↵ Murray Rosenblatt . Remarks on Some Nonparametric Estimates of a Density Function . The Annals of Mathematical Statistics , 27 ( 3 ): 832 – 837 , 1956 . Publisher: Institute of Mathematical Statistics . 2 OpenUrl CrossRef [21]. ↵ Chiranjibi Shah , MM Nabi , Simegnew Yihunie Alaba , Jack Prior , Ryan Caillouet , Matthew D Campbell , Farron Wallace , John E. Ball , and Robert Moorhead . A Zero Shot Detection Based Approach for Fish Species Recognition in Underwater Environments . In OCEANS 2023 - MTS/IEEE U.S. Gulf Coast , pages 1 – 7 , 2023 . ISSN: 0197-7385 . 2 [22]. ↵ Bart Van Der Sloot . Regulating the Synthetic Society: Generative AI, Legal Questions and Societal Challenges . Hart Publishing , 1 edition, 2024 . 2 [23]. ↵ Gaoang Wang , Jenq-Neng Hwang , Farron Wallace , and Craig Rose . Multi-Scale Fish Segmentation Refinement and Missing Shape Recovery . IEEE Access , 7 : 52836 – 52845 , 2019 . Conference Name: IEEE Access . 2 OpenUrl [24]. ↵ Yuntao Wang , Yanghe Pan , Miao Yan , Zhou Su , and Tom H. Luan . A Survey on ChatGPT: AI–Generated Contents, Challenges, and Solutions . IEEE Open Journal of the Computer Society , 4 : 280 – 302 , 2023 . Conference Name: IEEE Open Journal of the Computer Society . 2 OpenUrl [25]. ↵ Wenbo Zhang , Chaoyi Wu , and Zhenshan Bao . DPANet: Dual Pooling-aggregated Attention Network for fish segmentation . IET Computer Vision , 16 ( 1 ): 67 – 82 , 2022 . eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1049/cvi2.12065 . 2 OpenUrl [26]. ↵ Q. Zhu and J. Luo . Generative Pre-Trained Transformer for Design Concept Generation: An Exploration . Proceedings of the Design Society , 2 : 1825 – 1834 , 2022 . 2 OpenUrl View the discussion thread. Back to top Previous Next Posted July 10, 2025. Download PDF Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following FishProx: Proximate conspecific interaction of Atlantic salmon (Salmo salar) for behavioural analysis through instance segmentation masks Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share FishProx: Proximate conspecific interaction of Atlantic salmon ( Salmo salar ) for behavioural analysis through instance segmentation masks I-Hao Chen , Nabil Belbachir , Lars Ebbesson , Antonella Zanna Munthe-Kaas , Lars Erik Solberg , Gaute Alexander Nedberg Helberg , David Izquierdo-Gomez , Santhosh K. Kumaran , Jelena Kolarevic , Chris Noble bioRxiv 2025.07.07.659286; doi: https://doi.org/10.1101/2025.07.07.659286 Share This Article: Copy Citation Tools FishProx: Proximate conspecific interaction of Atlantic salmon ( Salmo salar ) for behavioural analysis through instance segmentation masks I-Hao Chen , Nabil Belbachir , Lars Ebbesson , Antonella Zanna Munthe-Kaas , Lars Erik Solberg , Gaute Alexander Nedberg Helberg , David Izquierdo-Gomez , Santhosh K. Kumaran , Jelena Kolarevic , Chris Noble bioRxiv 2025.07.07.659286; doi: https://doi.org/10.1101/2025.07.07.659286 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Animal Behavior and Cognition Subject Areas All Articles Animal Behavior and Cognition (7633) Biochemistry (17680) Bioengineering (13889) Bioinformatics (41927) Biophysics (21445) Cancer Biology (18585) Cell Biology (25491) Clinical Trials (138) Developmental Biology (13373) Ecology (19897) Epidemiology (2067) Evolutionary Biology (24308) Genetics (15606) Genomics (22496) Immunology (17736) Microbiology (40385) Molecular Biology (17175) Neuroscience (88583) Paleontology (666) Pathology (2830) Pharmacology and Toxicology (4822) Physiology (7641) Plant Biology (15149) Scientific Communication and Education (2045) Synthetic Biology (4293) Systems Biology (9822) Zoology (2271)