pith. sign in

arxiv: 2606.10940 · v1 · pith:PBDZ63FDnew · submitted 2026-06-09 · 💻 cs.CV · cs.AI· cs.LG

Democratising Camera Trap AI: An Open-Source Model for Detecting UK Mammals

Paul Fergus , Philip Stephens , Russell A. Hill , Lee Oliver , Katie Appleby , Sarah Beatham , Naomi Davies Walsh , Stuart Nixon
show 3 more authors
Naomi Matthews Chris Sutherland Kelly Hitchcock
This is my paper

Pith reviewed 2026-06-27 13:25 UTC · model grok-4.3

classification 💻 cs.CV cs.AIcs.LG
keywords camera trapobject detectionUK mammalsopen-source modelYOLObiodiversity monitoringartificial intelligenceecology
0
0 comments X

The pith

An open-source YOLO model detects 28 UK mammal and bird species in camera trap images with 0.984 mean average precision.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper releases a freely available object detection model trained to identify common UK mammals and birds in camera trap photographs. The model was built from over 48,000 labelled images collected over a decade from multiple sites. It reaches high accuracy on test images drawn from the same locations used in training. By providing the weights in an accessible format with desktop and real-time support, the work aims to let ecologists without machine learning expertise run the system locally instead of relying on commercial services. This directly addresses the barrier of proprietary AI tools that are often trained on non-UK fauna.

Core claim

The authors trained a YOLO26x detector on a dataset of 48,165 labelled instances covering 31 classes including 28 UK mammal and bird species plus humans, poles, and vehicles. On a held-out validation set the model attains a mean average precision of 0.984 at IoU 0.5 and 0.956 at IoU 0.5-0.95, with precision 0.988 and recall 0.965. On a further unseen test split, per-species confidence scores range from 0.96 to 0.99 and the false-negative rate is 0.17 percent, mostly in challenging night-time or occluded images. The trained weights are released in ONNX format under a non-commercial licence together with local desktop and real-time camera support.

What carries the argument

YOLO26x object detector trained on a class-stratified 80/10/10 split of the 48,165-instance UK camera trap dataset.

If this is right

  • Ecologists can run the detector locally on desktop computers without needing cloud services or machine learning expertise.
  • The model supports real-time processing from camera traps in the field.
  • Performance metrics hold on data from the training sites, providing a baseline for UK-specific fauna detection.
  • The non-commercial licence allows free use by conservation groups while restricting commercial exploitation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Performance on entirely new sites and camera setups remains untested and may require additional fine-tuning or data collection.
  • Integration with existing biodiversity databases could automate species occurrence mapping at scale.
  • Local processing reduces data transmission costs and privacy concerns associated with uploading images to commercial platforms.

Load-bearing premise

The high accuracy metrics were obtained on images from the same sites and cameras used to collect the training data.

What would settle it

Measuring the model's precision and recall on camera trap images collected from entirely new locations and camera models not represented in the training set.

Figures

Figures reproduced from arXiv: 2606.10940 by Chris Sutherland, Katie Appleby, Kelly Hitchcock, Lee Oliver, Naomi Davies Walsh, Naomi Matthews, Paul Fergus, Philip Stephens, Russell A. Hill, Sarah Beatham, Stuart Nixon.

Figure 1
Figure 1. Figure 1: Training dynamics over 120 epochs. The x-axis of every panel is the training epoch. Top row: training losses for the three components of the detector's objective, bounding-box regression (box_loss), species classification (cls_loss), and distribution focal loss (dfl_loss), which refines box-edge localisation. Bottom-left three panels: the same three losses computed on the held-out validation split (val/box… view at source ↗
read the original abstract

Camera traps have become a cornerstone of biodiversity monitoring, but the artificial intelligence that turns vast quantities of images into usable ecological data is often locked behind commercial platforms or trained on fauna that does not match that of the British Isles. In an attempt to remove barriers and increase uptake, we release an open-source object detection model for 31 classes, 28 common UK mammal and bird species, plus utility classes for humans, calibration poles, and vehicles, drawn from a curated dataset of 48,165 labelled instances assembled from multiple sites over a decade of operational deployment through Conservation AI and its successor, Trap Tracker. The model, a YOLO26x detector trained and tested on an 80/10/10 class-stratified split, achieves a mean Average Precision of 0.984 at Intersection over Union (IoU) of 0.5 (0.956 at IoU 0.5-0.95) on the held-out validation set, with precision 0.988 and recall 0.965. On an unseen held-out test split, mean per-species confidence ranged from 0.96 to 0.99 across the 31 classes, with a 0.17% false-negative rate concentrated in difficult night-time, distant, or occluded images. These metrics are from data from the same pool of sites and cameras as training, so performance at entirely new sites is left to future work. We release the trained weights in ONNX format under a non-commercial licence, with local desktop and real-time camera support, aimed explicitly at ecologists with no machine-learning experience. This release is a deliberate counterweight to the multiple paid for models that have developed over the last decade.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper presents an open-source YOLO26x object detection model trained to detect 31 classes (28 UK mammal/bird species plus humans, calibration poles, and vehicles) from camera-trap images. The training set comprises 48,165 labelled instances collected over a decade from multiple UK sites via Conservation AI/Trap Tracker. Using a class-stratified 80/10/10 split, the model reports mAP 0.984 at IoU=0.5 (0.956 at IoU 0.5-0.95), precision 0.988 and recall 0.965 on the held-out validation set; per-class confidence on a further unseen test split ranges 0.96-0.99 with 0.17% false-negative rate. The authors explicitly state that all reported figures are from the same site/camera distribution as training and defer evaluation on entirely new sites to future work. Trained weights are released in ONNX format under a non-commercial licence together with desktop and real-time camera support.

Significance. If the reported metrics hold under the stated distribution, the release supplies a practical, locally runnable alternative to commercial camera-trap classifiers for UK fauna, directly addressing the accessibility barrier noted in the abstract. The explicit scope limitation (same-site performance only) and the provision of reproducible weights constitute strengths that increase the utility of the contribution for ecologists without machine-learning expertise.

minor comments (2)
  1. [Abstract] §Abstract and §Methods: the text refers both to a 'held-out validation set' and a separate 'unseen held-out test split'; a single sentence clarifying the exact 80/10/10 partition (train/val/test) and which split supplies the mAP numbers versus the per-class confidence numbers would remove any ambiguity.
  2. [Results] Table 1 (or equivalent results table): per-class AP values are not shown; adding a supplementary table or figure with the 31 individual APs would allow readers to verify that the mean is not driven by a few dominant classes.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive review and recommendation to accept the manuscript.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper reports standard empirical machine-learning results: a YOLO26x model is trained on an 80/10/10 class-stratified split of a curated image dataset and evaluated directly on the held-out validation and test portions, yielding mAP, precision, and recall figures. No equations, first-principles derivations, or predictions are presented that reduce by construction to fitted parameters or self-referential definitions. The manuscript explicitly flags that the reported numbers apply only to the same site/camera pool and leaves generalization to new sites for future work, avoiding any over-claim. No load-bearing self-citations, uniqueness theorems, or ansatzes appear in the provided text. The central claim is therefore a direct measurement on held-out data and remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central performance claim rests on standard supervised object-detection assumptions plus the representativeness of the held-out split drawn from the original sites.

free parameters (1)
  • class-stratified 80/10/10 split
    The split ratio and stratification are chosen by the authors to create the reported validation and test sets.
axioms (1)
  • domain assumption Images from the same sites and cameras as training are a valid proxy for model evaluation
    All reported metrics are computed on held-out data from the original deployment sites.

pith-pipeline@v0.9.1-grok · 5876 in / 1284 out tokens · 22625 ms · 2026-06-27T13:25:46.480174+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

55 extracted references · 1 linked inside Pith

  1. [1]

    An evaluation of camera traps for inventorying large -and medium -sized terrestrial rainforest mammals,

    M. W. Tobler, S. E. Carrillo -Percastegui, R. Leite Pitman, R. Mares, and G. Powell, “An evaluation of camera traps for inventorying large -and medium -sized terrestrial rainforest mammals,” Animal conservation , vol. 11, no. 3, pp. 169 – 178, 2008

    2008

  2. [2]

    Camera-trapping for conservation: a guide to best-practices,

    O. R. Wearn and P. Glover -Kapfer, “Camera-trapping for conservation: a guide to best-practices,” WWF conservation technology series, vol. 1, no. 1, p. 181, 2017

    2017

  3. [3]

    An open standard for camera trap data,

    T. Forrester, T. O’Brien, E. Fegraus, P. A. Jansen, J. Palmer, R. Kays, J. Ahumada, B. Stern, and W. McShea, “An open standard for camera trap data,” Biodiversity data journal, no. 4, p. e10197, 2016

    2016

  4. [4]

    Software to facilitate and streamline camera trap data management: A review,

    S. Young, J. Rode -Margono, and R. Amin, “Software to facilitate and streamline camera trap data management: A review,” Ecology and Evolution , vol. 8, no. 19, pp. 9947 – 9957, 2018

    2018

  5. [5]

    Snapshot Serengeti, high -frequency annotated camera trap images of 40 mammalian species in an African savanna,

    A. Swanson, M. Kosmala, C. Lintott, R. Simpson, A. Smith, and C. Packer, “Snapshot Serengeti, high -frequency annotated camera trap images of 40 mammalian species in an African savanna,” Scientific data, vol. 2, no. 1, pp. 1–14, 2015

    2015

  6. [6]

    Camera settings and biome influence the accuracy of citizen science approaches to camera trap image classification,

    N. Egna, D. O’Connor, J. Stacy -Dawes, M. W. Tobler, N. Pilfold, K. Neilson, B. Simmons, E. O. Davis, M. Bowler, J. Fennessy, and others, “Camera settings and biome influence the accuracy of citizen science approaches to camera trap image classification,” Ecology and evolution , vol. 10, no. 21, pp. 11954–11965, 2020

    2020

  7. [7]

    Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning,

    M. S. Norouzzadeh, A. Nguyen, M. Kosmala, A. Swanson, M. S. Palmer, C. Packer, and J. Clune, “Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning,” Proceedings of the National Academy of Sciences, vol. 115, no. 25, pp. E5716– E5725, 2018

    2018

  8. [8]

    Machine learning to classify animal species in camera trap images: Applications in ecology,

    M. A. Tabak, M. S. Norouzzadeh, D. W. Wolfson, S. J. Sweeney, K. C. VerCauteren, N. P. Snow, J. M. Halseth, P. A. Di Salvo, J. S. Lewis, M. D. White, and others, “Machine learning to classify animal species in camera trap images: Applications in ecology,” Methods in Ecology and Evolution, vol. 10, no. 4, pp. 585–590, 2019

    2019

  9. [9]

    Improving the accessibility and transferability of machine learning algorithms for identification of animals in camera trap images: MLWIC2,

    M. A. Tabak, M. S. Norouzzadeh, D. W. Wolfson, E. J. Newton, R. K. Boughton, J. S. Ivan, E. A. Odell, E. S. Newkirk, R. Y. Conrey, J. Stenglein, and others, “Improving the accessibility and transferability of machine learning algorithms for identification of animals in camera trap images: MLWIC2,” Ecology and evolution, vol. 10, no. 19, pp. 10374–10383, 2020

    2020

  10. [10]

    Identifying animal species in camera trap images using deep learning and citizen science,

    M. Willi, R. T. Pitman, A. W. Cardoso, C. Locke, A. Swanson, A. Boyer, M. Veldthuis, and L. Fortson, “Identifying animal species in camera trap images using deep learning and citizen science,” Methods in Ecology and Evolution, vol. 10, no. 1, pp. 80–91, 2019

    2019

  11. [11]

    A deep active learning system for species identification and counting in camera trap images,

    M. S. Norouzzadeh, D. Morris, S. Beery, N. Joshi, N. Jojic, and J. Clune, “A deep active learning system for species identification and counting in camera trap images,” Methods in ecology and evolution, vol. 12, no. 1, pp. 150–161, 2021

    2021

  12. [12]

    Deep learning object detection methods for ecological camera trap data,

    S. Schneider, G. W. Taylor, and S. Kremer, “Deep learning object detection methods for ecological camera trap data,” in 2018 15th Conference on computer and robot vision (CRV), 2018, pp. 321–328

    2018

  13. [13]

    Long -tailed Species Recognition in the NACTI Wildlife Dataset,

    Z. Liu and T. Burghardt, “Long -tailed Species Recognition in the NACTI Wildlife Dataset,” arXiv preprint arXiv:2510.21657, 2025

    arXiv 2025

  14. [14]

    Recognition in terra incognita,

    S. Beery, G. Van Horn, and P. Perona, “Recognition in terra incognita,” in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 456–473

    2018

  15. [15]

    Robust ecological analysis of camera trap data labelled by a machine learning model,

    R. C. Whytock, J. wieewski, J. A. Zwerts, T. Bara -Slupski, A. F. Koumba Pambo, M. Rogala, L. Bahaa -el-din, K. Boekee, S. Brittain, A. W. Cardoso, and others, “Robust ecological analysis of camera trap data labelled by a machine learning model,” Methods in Ecology and Evolution, vol. 12, no. 6, pp. 1080–1092, 2021

    2021

  16. [16]

    Conservation AI: Live stream analysis for the detection of endangered species using convolutional neural networks and drone technology,

    C. Chalmers, P. Fergus, S. Wich, and A. C. Montanez, “Conservation AI: Live stream analysis for the detection of endangered species using convolutional neural networks and drone technology,” arXiv preprint arXiv:1910.07360, 2019

    arXiv 1910

  17. [17]

    Harnessing artificial intelligence for wildlife conservation,

    P. Fergus, C. Chalmers, S. Longmore, and S. Wich, “Harnessing artificial intelligence for wildlife conservation,” Conservation, vol. 4, no. 4, pp. 685 –702, 2024

    2024

  18. [18]

    Trap Tracker: AI -Driven Wildlife Monitoring for Conservation

    T. Tracker, “Trap Tracker: AI -Driven Wildlife Monitoring for Conservation.” 2026

    2026

  19. [19]

    Efficient pipeline for camera trap image review,

    S. Beery, D. Morris, and S. Yang, “Efficient pipeline for camera trap image review,” arXiv preprint arXiv:1907.06772, 2019

    Pith/arXiv arXiv 1907

  20. [20]

    Camera Trap AI: SpeciesNet for Camera Trap Image Classification

    T. Gadot and others, “Camera Trap AI: SpeciesNet for Camera Trap Image Classification.” Google AI for Nature and Wildlife Insights, 2024

    2024

  21. [21]

    The megadetector: Large -scale deployment of computer vision for conservation and biodiversity monitoring,

    S. Beery, “The megadetector: Large -scale deployment of computer vision for conservation and biodiversity monitoring,” California Institute of Technology, Pasadena, CA, USA, 2023

    2023

  22. [22]

    Wildlife insights: A platform to maximize the potential of camera trap and other passive sensor wildlife data for the planet ,

    J. A. Ahumada, E. Fegraus, T. Birch, N. Flores, R. Kays, T. G. O’Brien, J. Palmer, S. Schuttler, J. Y. Zhao, W. Jetz, and others, “Wildlife insights: A platform to maximize the potential of camera trap and other passive sensor wildlife data for the planet ,” Environmental Conservation, vol. 47, no. 1, pp. 1–6, 2020

    2020

  23. [23]

    The DeepFaune initiative: a collaborative effort towards the automatic identification of European fauna in camera trap images ,

    N. Rigoudy, G. Dussert, A. Benyoub, A. Besnard, C. Birck, J. Boyer, Y. Bollet, Y. Bunz, G. Caussimont, E. Chetouane, and others, “The DeepFaune initiative: a collaborative effort towards the automatic identification of European fauna in camera trap images ,” European Journal of Wildlife Research, vol. 69, no. 6, p. 113, 2023

    2023

  24. [24]

    cameratrapR: An R package for estimating animal density using camera trapping data,

    X. Li, H. Tian, Z. Piao, G. Wang, Z. Xiao, Y. Sun, E. Gao, and M. Holyoak, “cameratrapR: An R package for estimating animal density using camera trapping data,” Ecological Informatics, vol. 69, p. 101597, 2022

    2022

  25. [25]

    Camelot –intuitive software for camera trap data management,

    H. Hendry and C. Mann, “Camelot –intuitive software for camera trap data management,” BioRxiv, p. 203216, 2017

    2017

  26. [26]

    AddaxAI: Simplifying camera trap image analysis with AI

    A. D. Science, “AddaxAI: Simplifying camera trap image analysis with AI.” 2026

    2026

  27. [27]

    Pytorch -wildlife: A collaborative deep learning framework for conservation,

    A. Hernandez, Z. Miao, L. Vargas, S. Beery, R. Dodhia, P. Arbelaez, and J. M. L. Ferres, “Pytorch -wildlife: A collaborative deep learning framework for conservation,” arXiv preprint arXiv:2405.12930, 2024

    arXiv 2024

  28. [28]

    DeepForestVision: Automated wildlife identification for camera traps of African tropical forests,

    H. Magaldi, R. Cornette, J. J. Tibesigwa, R. Katumba, H. Rugonge, B. Amarasekaran, N. Anderson, N. Cappelle, A. W. Cardoso, D. Cornélis, and others, “DeepForestVision: Automated wildlife identification for camera traps of African tropical forests,” Ecological Solutions and Evidence, vol. 6, no. 4, p. e70167, 2025

    2025

  29. [29]

    Yolo -animal: An efficient wildlife detection network based on improved yolov5,

    D. Ma and J. Yang, “Yolo -animal: An efficient wildlife detection network based on improved yolov5,” in 2022 International Conference on Image Processing, Computer Vision and Machine Learning (ICICML) , 2022, pp. 464 – 468

    2022

  30. [30]

    An evaluation of platforms for processing camera- trap data using artificial intelligence,

    J. Vélez, W. McShea, H. Shamon, P. J. Castiblanco - Camacho, M. A. Tabak, C. Chalmers, P. Fergus, and J. Fieberg, “An evaluation of platforms for processing camera- trap data using artificial intelligence,” Methods in Ecology and Evolution, vol. 14, no. 2, pp. 459–477, 2023

    2023

  31. [31]

    Animal Detect: Easy camera trap image analysis for wildlife teams

    A. Detect, “Animal Detect: Easy camera trap image analysis for wildlife teams.” 2026

    2026

  32. [32]

    Agouti: A platform for processing and archiving of camera trap images,

    J. Casaer, T. Milotic, Y. Liefting, P. Desmet, and P. Jansen, “Agouti: A platform for processing and archiving of camera trap images,” Biodiversity Information Science and Standards, 2019

    2019

  33. [33]

    Global disparity of camera trap research allocation and defaunation risk of terrestrial mammals,

    B. Mugerwa, J. Niedballa, A. Planillo, D. Sheil, S. Kramer- Schadt, and A. Wilting, “Global disparity of camera trap research allocation and defaunation risk of terrestrial mammals,” Remote Sensing in Ecology and Conservation , vol. 10, no. 1, pp. 121–136, 2024

    2024

  34. [34]

    People’s Trust for Endangered Species (PTES)

    P. T. for Endangered Species, “People’s Trust for Endangered Species (PTES).”

  35. [35]

    The Mammal Society: Science -led conservation of British mammals

    T. M. Society, “The Mammal Society: Science -led conservation of British mammals.”

  36. [36]

    National Biodiversity Network (NBN)

    N. B. N. Trust, “National Biodiversity Network (NBN).”

  37. [37]

    MammalWeb: A citizen science platform for monitoring wild mammals using camera traps

    MammalWeb, “MammalWeb: A citizen science platform for monitoring wild mammals using camera traps.”

  38. [38]

    Large -scale mammal monitoring: The potential of a citizen science camera-trapping project in the United Kingdom,

    P.-Y. Hsing, R. A. Hill, G. C. Smith, S. Bradley, S. E. Green, V. T. Kent, S. S. Mason, J. Rees, M. J. Whittingham, J. Cokill, and others, “Large -scale mammal monitoring: The potential of a citizen science camera-trapping project in the United Kingdom,” Ecological Solutions and Evidence, vol. 3, no. 4, p. e12180, 2022

    2022

  39. [39]

    State of Nature 2023 Partners

    S. of Nature Partnership, “State of Nature 2023 Partners.” 2023

    2023

  40. [40]

    Environment Act 2021

    U. Parliament, “Environment Act 2021.” 2021

    2021

  41. [41]

    UK Biodiversity Indicators: Priority Species — Relative Abundance

    J. N. C. Committee, “UK Biodiversity Indicators: Priority Species — Relative Abundance.” JNCC on behalf of the Department for Environment, Food and Rural Affairs (Defra), 2024

    2024

  42. [42]

    Indicators of Species Abundance in England

    F. Department for Environment and R. Affairs, “Indicators of Species Abundance in England.” 2024

    2024

  43. [43]

    Kunming -montreal global biodiversity framework,

    U. CBD, “Kunming -montreal global biodiversity framework,” in Fifteenth meeting of the Conference of the Parties to the Convention on Biological Diversity (Part Two) Decision, 2022, vol. 15, no. 4

    2022

  44. [44]

    The Kunming-Montreal Global Biodiversity Framework: what it does and does not do, and how to improve it,

    A. C. Hughes and R. E. Grumbine, “The Kunming-Montreal Global Biodiversity Framework: what it does and does not do, and how to improve it,” Frontiers in Environmental Science, vol. 11, p. 1281536, 2023

    2023

  45. [45]

    Automated detection of European wild mammal species in camera trap images with an existing and pre -trained computer vision model,

    C. Carl, F. Schönfeld, I. Profft, A. Klamm, and D. Landgraf, “Automated detection of European wild mammal species in camera trap images with an existing and pre -trained computer vision model,” European journal of wildlife research, vol. 66, no. 4, p. 62, 2020

    2020

  46. [46]

    Increasing citizen scientist accuracy with artificial intelligence on UK camera- trap data,

    C. Sharpe, R. Hill, H. Chappell, S. Green, K. Holden, P. Fergus, C. Chalmers, and P. Stephens, “Increasing citizen scientist accuracy with artificial intelligence on UK camera- trap data,” Remote Sensing in Ecology and Conservation , vol. 11, no. 6, pp. 641–655, 2025

    2025

  47. [47]

    Optimizing the Accuracy and Efficiency of Camera Trap Image Analysis: Evaluating AI Model Performance and a Semi -Automated Workflow,

    K. Hitchcock, S. Tollington, R. W. Yarnell, L. J. Williams, K. Hamill, and P. Fergus, “Optimizing the Accuracy and Efficiency of Camera Trap Image Analysis: Evaluating AI Model Performance and a Semi -Automated Workflow,” Remote Sensing, vol. 18, no. 3, p. 502, 2026

    2026

  48. [48]

    Comparison of a Computer Vision Model to a Human Observer in Detecting African Mammals in Camera Trap Images within a Safari Park,

    N. Davies Walsh, C. Chalmers, P. Fergus, S. Longmore, B. Johnson, and S. Wich, “Comparison of a Computer Vision Model to a Human Observer in Detecting African Mammals in Camera Trap Images within a Safari Park,” bioRxiv, pp. 2025–6, 2025

    2025

  49. [49]

    Towards context -rich automated biodiversity assessments: deriving AI-powered insights from camera trap data,

    P. Fergus, C. Chalmers, N. Matthews, S. Nixon, A. Burger, O. Hartley, C. Sutherland, X. Lambin, S. Longmore, and S. Wich, “Towards context -rich automated biodiversity assessments: deriving AI-powered insights from camera trap data,” Sensors, vol. 24, no. 24, p. 8122, 2024

    2024

  50. [50]

    Empirical evidence that diversionary feeding increases productivity in ground -nesting birds,

    J. A. Bamber, C. Sutherland, K. Kortland, and X. Lambin, “Empirical evidence that diversionary feeding increases productivity in ground -nesting birds,” Proceedings of the Royal Society B: Biological Sciences , vol. 292, no. 2049, 2025

    2049

  51. [51]

    YOLO26: key architectural enhancements and performance benchmarking for real -time object detection,

    R. Sapkota, R. H. Cheppally, A. Sharda, and M. Karkee, “YOLO26: key architectural enhancements and performance benchmarking for real -time object detection,” arXiv preprint arXiv:2509.25164, 2025

    arXiv 2025

  52. [52]

    Lessons and Open Questions from a Unified Study of Camera -Trap Species Recognition Over Time,

    S. Jeon, H. Tian, L. Wang, Z. Mai, V. Bakshi, J. Hou, P. Zhang, A. Chowdhury, J. Gu, and W.-L. Chao, “Lessons and Open Questions from a Unified Study of Camera -Trap Species Recognition Over Time,” arXiv preprint arXiv:2603.20509, 2026

    arXiv 2026

  53. [53]

    Empowering wildlife guardians: An equitable digital stewardship and reward system for biodiversity conservation using deep learni ng and 3/4G camera traps,

    P. Fergus, C. Chalmers, S. Longmore, S. Wich, C. Warmenhove, J. Swart, T. Ngongwane, A. Burger, J. Ledgard, and E. Meijaard, “Empowering wildlife guardians: An equitable digital stewardship and reward system for biodiversity conservation using deep learni ng and 3/4G camera traps,” Remote sensing , vol. 15, no. 11, p. 2730, 2023

    2023

  54. [54]

    Removing human bottlenecks in bird classification using camera trap images and deep learning,

    C. Chalmers, P. Fergus, S. Wich, S. N. Longmore, N. D. Walsh, P. A. Stephens, C. Sutherland, N. Matthews, J. Mudde, and A. Nuseibeh, “Removing human bottlenecks in bird classification using camera trap images and deep learning,” Remote Sensing, vol. 15, no. 10, p. 2638, 2023

    2023

  55. [55]

    AI-driven real-time monitoring of ground -nesting birds: a case study on curlew detection using YOLOv10,

    C. Chalmers, P. Fergus, S. Wich, S. N. Longmore, N. D. Walsh, L. Oliver, J. Warrington, J. Quinlan, and K. Appleby, “AI-driven real-time monitoring of ground -nesting birds: a case study on curlew detection using YOLOv10,” Remote Sensing, vol. 17, no. 5, p. 769, 2025. APPENDIX A: SAMPLE TRAINING MOSAICS - SHOWING CONDITIONS THAT ANY UK CAMERA TRAP DEPLOYM...

    2025