Datasets with annotated wildlife in drone/aerial images
July 23, 2026 · View on GitHub
Permalink to this page:
http://lila.science/aerialdata
Contents
- Overview
- Publicly available datasets
- Datasets added recently that I haven't had a chance to dig into yet
- Datasets available by request
- Publicly-available models for wildlife detection in drone/aerial images
- Platforms/systems for wildlife detection in drone/aerial images
- OSS repos about wildlife detection in drone/aerial images
Overview
This is a list of datasets with annotated aerial/drone/satellite imagery for wildlife surveys. I've tried to collect basic standardized metadata for each dataset, and to provide sample code in a similar format for each dataset that can match annotations to images and render a sample image. The goal of this exercise was to make it easier to assess whether existing datasets could be useful for training new models, e.g. if you want to find birds in images with a particular gestalt, we wanted to make it easier for you to get the gestalt of existing datasets, and have a starting point for parsing that data.
Everything listed here is also either listed on LILA's list of other conservation datasets or is on LILA.
This list is maintained by Dan Morris, but it began life as a thread on the AI for Conservation Slack, and was initially assembled with help from Zhongqi Miao, Kalindi Fonda, Aakash Gupta, Josh Veitch-Michaelis, and Ed Bayes.
Email Dan if anything seems off, or if you know of datasets I'm missing.
As a bonus, this page is also a temporary holding place for a list of models for wildlife detection in aerial/drone imagery.
Publicly available datasets
Improving the precision and accuracy of animal population estimates with aerial image object detection
Aerial images with 4305 bounding boxes on zebra, giraffe, and elephants
Eikelboom JA, Wind J, van de Ven E, Kenana LM, Schroder B, de Knegt HJ, van Langevelde F, Prins HH. Improving the precision and accuracy of animal population estimates with aerial image object detection. Methods in Ecology and Evolution. 2019 Nov;10(11):1875-87.
- 5.7 GB, downloadable via http from 4TU (download link)
- Metadata in csv format
- Categories: elephant, zebra, giraffe
- Vehicle type: plane
- Image information: 561 RGB images
- Annotation information: 4305 boxes
- Typical animal size in pixels: 50
- License: CC0
- Code to render sample annotated image: preview-eikelboom-savanna.py
- Shortcode: eikelboom-savanna
UAV-derived waterfowl thermal imagery dataset
8976 bounding boxes on waterfowl in UAV-derived thermal images
Hu Q, Smith J, Woldt W, Tang Z. UAV-derived waterfowl thermal imagery dataset. Mendeley Data, V4. 2021.
- 4.1 GB, downloadable via http from Mendeley Data (download link)
- Metadata in csv format
- Categories: waterfowl
- Vehicle type: plane
- Image information: 541 thermal images (unannotated RGB images included as a visual reference)
- Annotation information: 8976 boxes
- Typical animal size in pixels: 7
- License: CC BY 4.0
- Code to render sample annotated image: preview-hu-thermal.py
- Shortcode: hu-thermal
Drones count wildlife more accurately and precisely than humans
Images of 10 life-sized, replica seabird colonies containing a known number of fake birds taken from four different heights (30m, 60m, 90m and 120m)
Hodgson JC, Mott R, Baylis SM, Pham TT, Wotherspoon S, Kilpatrick AD, Raja Segaran R, Reid I, Terauds A, Koh LP. Drones count wildlife more accurately and precisely than humans. Methods in Ecology and Evolution. 2018 May;9(5):1160-7.
- 50 MB, downloadable via http from Dryad (download link)
- Metadata in csv format
- Categories: fake seabirds
- Vehicle type: drone
- Image information: 40 RGB images
- Annotation information: 1560 counts
- Typical animal size in pixels: variable
- License: CC0
- Shortcode: hodgson-counts
Counting animals in aerial images with a density map estimation model
137365 point annotations on penguins in RGB UAV images
Qian Y, Humphries G, Trathan P, Lowther A, Donovan C. Counting animals in aerial images with a density map estimation model [Data set]. 2023.
- 300 MB, downloadable via http from Zenodo (download link)
- Metadata in json format (LabelBox standard)
- Categories: brush-tailed penguins
- Vehicle type: plane
- Image information: 753 RGB images (orthorectified)
- Annotation information: 137365 points
- Typical animal size in pixels: 30
- License: CC0
- Code to render sample annotated image: preview-qian-penguins.py
- Shortcode: qian-penguins
Data from: A convolutional neural network for detecting sea turtles in drone imagery
1902 point annotations on sea turtles in drone images
Gray PC, Fleishman AB, Klein DJ, McKown MW, Bezy VS, Lohmann KJ, Johnston DW. A convolutional neural network for detecting sea turtles in drone imagery. Methods in Ecology and Evolution. 2019 Mar;10(3):345-55.
- 7.24 GB, downloadable via http from Zenodo (download link)
- Metadata in csv format
- Categories: olive ridley turtle
- Vehicle type: drone
- Image information: 1059 NIR images (false color NIR rendered to RGB)
- Annotation information: 1902 points
- Typical animal size in pixels: 10
- License: CC0
- Code to render sample annotated image: preview-gray-turtles.py
- Shortcode: gray-turtles
The Aerial Elephant Dataset
15581 point annotations on elephants in aerial images
Naude J, Joubert D. The Aerial Elephant Dataset: A New Public Benchmark for Aerial Object Detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops 2019 (pp. 48-55).
- 16.3 GB, downloadable via http from Zenodo (download link)
- Metadata in csv format
- Categories: elephant
- Vehicle type: drone
- Image information: 2074 RGB images
- Annotation information: 15581 points
- Typical animal size in pixels: 75
- License: CC0
- Code to render sample annotated image: preview-aerial-elephants.py
- Shortcode: aerial-elephants
A global model of bird detection in high resolution airborne images using computer vision
386638 box annotations on 23765 drone images from 13 ecosystems
Weinstein BG, Garner L, Saccomanno VR, Steinkraus A, Ortega A, Brush K, Yenni G, McKellar AE, Converse R, Lippitt CD, Wegmann A. A general deep learning model for bird detection in high-resolution airborne imagery. Ecological Applications. 2022 Dec:e2694.
- 29 GB, downloadable via http from Zenodo (download link)
- Metadata in csv format
- Categories: bird
- Vehicle type: variable
- Image information: 23765 RGB images
- Annotation information: 386638 boxes
- Typical animal size in pixels: 35
- License: CC BY 4.0
- Code to render sample annotated image: preview-weinstein-birds.py
- Shortcode: weinstein-birds
Aerial Photo Imagery from Fall Waterfowl Surveys, Izembek Lagoon, Alaska, 2017-2019
631349 point annotations on waterfowl on 110,067 aerial images
Weiser EL, Flint PL, Marks DK, Shults BS, Wilson HM, Thompson SJ, Fischer JB. Aerial photo imagery from fall waterfowl surveys, Izembek Lagoon, Alaska, 2017-2019: U.S. Geological Survey data release. 2022.
- 1.82 TB, downloadable via http from USGS (download link)
- A slightly-more-curated version of this dataset, minus many of the blank images, is hosted on LILA. This version is 124GB.
- Metadata for the original version is in csv, json format (CountThings format); metadata for the LILA version is in COCO format.
- Categories: brant goose, emperor goose, canada goose, gull, other
- Vehicle type: plane
- Image information: 110667 RGB images
- Annotation information: 631349 points
- Typical animal size in pixels: 50
- License: unspecified, but public domain implied (USGS source)
- Code to render sample annotated image from the original dataset: preview-weiser-waterfowl.py
- Code to render sample annotated image from the LILA dataset: preview-weiser-waterfowl-lila.py
- Shortcode: weiser-waterfowl-lila
Data from: Drones and deep learning produce accurate and efficient monitoring of large-scale seabird colonies
44966 bounding boxes on drone images of black-browed albatrosses and southern rockhopper penguins
Hayes MC, Gray PC, Harris G, Sedgwick WC, Crawford VD, Chazal N, Crofts S, Johnston DW. Data from: Drones and deep learning produce accurate and efficient monitoring of large-scale seabird colonies. Duke Research Repository. doi. 2020;10:r4dn45v9g.
- 20.5 GB, downloadable via Globus from Globus (download link)
- Metadata in csv format
- Categories: black-browed albatross, southern rockhopper penguin
- Vehicle type: drone
- Image information: 3947 RGB images
- Annotation information: 44966 boxes
- Typical animal size in pixels: 300
- License: CC0
- Code to render sample annotated image: preview-hayes-seabirds.py
- Shortcode: hayes-seabirds
Cattle detection and counting in UAV images based on convolutional neural networks
1919 bounding boxes on cattle in drone images, with individual IDs
Shao W, Kawakami R, Yoshihashi R, You S, Kawase H, Naemura T. Cattle detection and counting in UAV images based on convolutional neural networks. International Journal of Remote Sensing. 2020 Jan 2;41(1):31-52.
- 3.2 GB, downloadable via http from University of Tokyo (download link)
- Metadata in txt format
- Categories: cattle
- Vehicle type: drone
- Image information: 663 RGB images
- Annotation information: 1919 boxes
- Typical animal size in pixels: 90
- License: unspecified
- Code to render sample annotated image: preview-shao-cattle.py
- Shortcode: shao-cattle
NOAA Fisheries Steller Sea Lion Population Count
948 aerial images of sea lions with counts for each image
- 103 GB, downloadable via http or torrent from Kaggle (download link)
- Metadata in csv format
- Categories: Steller sea lion
- Vehicle type: plane
- Image information: 948 RGB images
- Annotation information: 948 counts
- Typical animal size in pixels: 75
- License: unspecified (public domain implied, NOAA source)
- Code to render sample annotated image: preview-steller-sea-lion-count.py
- Shortcode: steller-sea-lion-count
Right Whale Recognition
4544 images of right whales with individual IDs
- 10 GB, downloadable via http from Kaggle (download link)
- Metadata in csv format
- Categories: right whale
- Vehicle type: helicopter
- Image information: 11468 RGB images
- Annotation information: 4544 individual IDs
- Typical animal size in pixels: 1500
- License: unspecified (public domain implied, NOAA source)
- Code to render sample annotated image: preview-right-whale-recognition.py
- Shortcode: right-whale-recognition
NOAA Arctic Seals 2019
Around 14000 bounding boxes on seals in 44185 color/thermal image pairs
Alaska Fisheries Science Center, 2021: A Dataset for Machine Learning Algorithm Development.
- 1 TB, downloadable via azcopy from LILA (download link)
- Metadata in csv format
- Categories: ringed_seal, ringed_pup, unknown_seal, bearded_pup, bearded_seal, unknown_pup
- Vehicle type: plane
- Image information: 44185 RGB+IR images (RGB and IR are different resolutions, but registered well, and annotations for every animal are provided for both images)
- Annotation information: 14311 boxes
- Typical animal size in pixels: 55
- License: CDLA-permissive
- Code to render sample annotated image: preview-noaa-arctic-seals.py
- Shortcode: noaa-arctic-seals
Aerial Seabirds West Africa
High-resolution aerial RGB imagery obtained from a census of breeding seabirds in West Africa in 2019, with 21516 point annotations on seabirds
Kellenberger B, Veen T, Folmer E, Tuia D. 21,000 birds in 4.5 h: efficient large-scale seabird detection with machine learning. Remote Sensing in Ecology and Conservation. 2021.
- 2.2 GB, downloadable via http or azcopy from LILA (download link)
- Metadata in csv format
- Categories: great white pelican, royal tern, caspian tern, slender-billed gull, gray-headed gull, great cormorant
- Vehicle type: plane
- Image information: single aerial orthomosaic RGB images
- Annotation information: 21516 points
- Typical animal size in pixels: 30
- License: CDLA-permissive
- Code to render sample annotated image: preview-aerial-seabirds-west-africa.py
- Shortcode: aerial-seabirds-west-africa
Conservation Drones
Thermal drone videos with 166221 boxes and object IDs on humans, elephants, and several other animals
Bondi E, Jain R, Aggrawal P, Anand S, Hannaford R, Kapoor A, Piavis J, Shah S, Joppa L, Dilkina B, Tambe M. BIRDSAI: A Dataset for Detection and Tracking in Aerial Thermal Infrared Videos.
- 3.7 GB, downloadable via http or azcopy from LILA (download link)
- Metadata in csv format (MOT standard)
- Categories: human, elephant, giraffe, lion, dog
- Vehicle type: drone
- Image information: 61994 frames from 48 thermal videos
- Annotation information: 166221 boxes
- Typical animal size in pixels: 35
- License: CDLA-permissive
- Code to render sample annotated image: preview-conservation-drones.py
- Shortcode: conservation-drones
BuckTales: A multi-UAV dataset for multi-object tracking and re-identification of wild antelopes
This is really three datasets in one: a detection dataset, a re-ID dataset, and a tracking dataset, all using an overlapping set of videos, specifically UAV images over a herd of blackbuck. The detection dataset is 320 images with 18.4k boxes, the tracking dataset is 1.2M boxes over 12 video sequences (it's not explicitly stated, but I think those boxes are generated via a detector trained on the detection dataset), and the Re-ID dataset labels 730 individuals.
Naik H, Yang J, Das D, Crofoot MC, Rathore A, Sridhar VH. BuckTales: A multi-UAV dataset for multi-object tracking and re-identification of wild antelopes. Advances in Neural Information Processing Systems. 2024 Dec 16;37:81992-2009.
- 80GB, downloadable via http from Edmond (download link)
- Metadata in COCO and YOLO format (for the detection dataset) and MOT format (for the tracking dataset)
- Categories: drone, bird, unknown, shadow, female blackbuck, male blackbuck
- Vehicle type: drone
- License: CC BY-SA 4.0
- Annotation information: 18.4k boxes in the detection dataset
- Image information: 320 images
- Annotation information: 21,069 boxes
- Typical animal size in pixels: 40
- Code to render sample annotated image: preview-naik-bucktales.py
- Shortcode: naik-bucktales
DAZZLE: Drone-Acquired Zebra Data for Large-scale Ecology Research
Oblique aerial videos of zebras with 162931 bounding boxes and behavioral labels (standing, grazing, etc.).
Price E, Khandelwal PC, Rubenstein DI, Ahmad A. A Framework for Fast, Large-scale, Semi-Automatic Inference of Animal Behavior from Monocular Videos. bioRxiv. 2023:2023-07.
- 96GB, downloadable via http from Darus(download link)
- More information available on the dataset home page
- Metadata in json format (Labelme standard)
- Categories: zebra, person, vehicle
- Vehicle type: plane
- Image information: 30,6026 video frames (~4k resolution)
- Annotation information: 162931 boxes (4387 fully manual, 158544 semi-automated) with behavior labels
- Typical animal size in pixels: 134
- License: unspecified
- Code to render sample annotated image: preview-price-zebras.py
- Shortcode: price-zebras
Quantifying the movement, behaviour and environmental context of group-living animals using drones and computer vision
Drone images of ungulates and geladas with 40532 bounding boxes.
Koger B, Deshpande A, Kerby JT, Graving JM, Costelloe BR, Couzin ID. Quantifying the movement, behaviour and environmental context of group-living animals using drones and computer vision. Journal of Animal Ecology. 2023 Mar 21.
- 65GB, downloadable via http from Edmond (download link)
- Metadata in json format (COCO standard)
- Categories: zebra, gazelle, waterbuck, buffalo, other, gelada, human
- Vehicle type: drone
- Image information: 1982 drone images
- Annotation information: 40532 boxes
- Typical animal size in pixels: 56
- License: CC0
- Code to render sample annotated image: preview-koger-drones.py
- Shortcode: koger-drones
Data from "Deep object detection for waterbird monitoring using aerial imagery"
Kabra K, Xiong A, Li W, Luo M, Lu W, Yu T, Yu J, Singh D, Garcia R, Tang M, Arnold H. Deep object detection for waterbird monitoring using aerial imagery. In 2022 21st IEEE International Conference on Machine Learning and Applications (ICMLA) 2022 Dec 12 (pp. 455-460). IEEE.
- 3.7GB, downloadable from Google Drive (download link)
- Metadata in .csv format
- Categories: list of categories
- Vehicle type: drone
- Image information: 200 drone images
- Annotation information: 23078 boxes, species and age (juvenile/adult) information for each box
- Typical animal size in pixels: 88.9
- License: unspecified
- Code to render sample annotated image: preview-kabra-birds.py
- Shortcode: kabra-birds
KABR: In-Situ Dataset for Kenyan Animal Behavior Recognition from Drone Videos
10 hours of UAV video from Kenyan savanna, with behavior labels.
Kholiavchenko M, Kline J, Ramirez M, Stevens S, Sheets A, Babu R, Banerji N, Campolongo E, Thompson M, Van Tiel N, Miliko J. KABR: In-Situ Dataset for Kenyan Animal Behavior Recognition From Drone Videos. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision 2024.
- 54GB, downloadable via http from Google Drive (download link) (Hugging Face link)
- Metadata in space-delimited table format
- Categories: (walk, graze, browse, head up, auto-groom, trot, run, occluded) for (giraffe, zebra)
- Vehicle type: drone
- Image information: 130366 videos each following a single individual, provided as ~1.1M JPGs
- Annotation information: behavior labels for individual frames
- Typical animal size in pixels: 100
- License: unspecified
- Code to render sample annotated image: preview-kabr-behavior.py
- Shortcode: kabr-behavior
MMLA-OPC
~29k frames with boxes on zebras in UAV images, collected from Ol Pejeta Conservancy in Kenya.
Kline J, Nguyen Ngoc D, Duncan H, Rondaeu Saint-Jean C, Maalouf G, Juma B, Kilwaya A, Vuyiya B, Irungu M, Njoroge W, Mutisya S, Guerin D, Costelloe B, Pastucha E, Hermansen J, Kjeld J, Watson M, Richardson T, Schultz Lundquist UP. MMLA Ol Pejeta Conservancy (OPC) Dataset, 2025.
- 64GB, downloadable from Hugging Face
- Metadata in YOLO format
- Categories: zebra
- Vehicle type: drone
- Image information: 29,268 RGB frames
- Annotation information: ~163k bounding boxes
- Typical animal size in pixels
- License: CC 1.0
- Code to render sample annotated image: preview-mmla-opc.py
- Shortcode: mmla-opc
MMLA-Wilds
~8k frames with boxes on onagers (~45k), giraffe (~7k), zebra (~2.5k) and wild dogs (14) in UAV images, collected at The Wilds Conservation Center in Ohio.
Kline J, Zhong A, Yablok J. MMLA The Wilds Dataset, 2025.
- 21GB, downloadable from Hugging Face
- Metadata in YOLO format
- Categories: zebra, giraffe, onager, dog
- Vehicle type: drone
- Image information: 8,009 RGB frames
- Annotation information: bounding boxes
- Typical animal size in pixels: 1000px
- License: CC 1.0
- Code to render sample annotated image: preview-mmla-wilds.py
- Shortcode: mmla-wilds
MMLA-Mpala
~130k frames with boxes on giraffe (~122k) and zebra (~495k) in UAV images, collected at Mpala Research Center in Kenya.
Kline J, Kholiavchenko M, Zhong Alison, Ramirez M, Stevens S, Van Tiel N, Campolongo E, Thompson M, Ramesh Babu R, Banerji N, Sheets A, Magersupp M, Balasubramaniam S, Duporge I, Miliko J, Rosser N, Stewart CV, Berger-Wolf T, Rubenstein DI.
MMLA Mpala Dataset, 2025.
- 490GB, downloadable from Hugging Face
- Metadata in YOLO format
- Categories: zebra, giraffe
- Vehicle type: drone
- Image information: 130,102 images
- Annotation information: ~617k bounding boxes
- Typical animal size in pixels: 1138px
- License: CC 1.0
- Code to render sample annotated image: preview-mmla-mpala.py
- Shortcode: mmla-mpala
WAID: Wildlife Aerial Images from Drone
Mou C, Liu T, Zhu C, Cui X. Waid: A large-scale dataset for wildlife detection with drones. Applied Sciences. 2023 Sep 17;13(18):10397.IEEE/CVF Winter Conference on Applications of Computer Vision 2024.
- 1.5GB, downloadable from GitHub (as in, literally from GitHub) (download link)
- Metadata in YOLO format
- Categories: sheep, cattle, seal, camel, kiang, zebra (some images also include un-annotated birds)
- Vehicle type: drone
- Image information: 14,366 images, typically 640x640
- Annotation information: boxes
- Typical animal size in pixels: 166
- License: unspecified
- Code to render sample annotated image: preview-waid-drones.py
- Shortcode: waid-drones
UAS Imagery of Migratory Waterfowl at New Mexico Wildlife Refuges
Converse RC, Lippitt CD, Sesnie SE, Harris GM, Butler MG, Stewart DR. Observer variability in manual-visual interpretation of UAS imagery of wildlife, with insights for deep learning applications. In review.
- 322MB, downloadable from LILA (download link)
- Metadata in COCO .json format
- Categories: canada goose, sandhill crane, mallard, northern pintail, american wigeon, teal, gadwall, northern shoveler, other
- Vehicle type: drone
- Image information: 12 images if ~5kx4k, 356 images of 684x521
- Annotation information: 2243 consensus boxes
- Typical animal size in pixels: 50
- License: CC-BY-NC 4.0
- Code to render sample annotated image: preview-nm-waterfowl.py
- Shortcode: nm-waterfowl
Multispecies detection and identification of African mammals
Delplanque A, Foucher S, Lejeune P, Linchant J, Théau J. Multispecies detection and identification of African mammals in aerial imagery using convolutional neural networks. Remote Sensing in Ecology and Conservation. 2022 Apr;8(2):166-79.
- 12GB, downloadable from Liege University Dataverse (download link)
- Metadata in COCO .json format
- Categories: alcelaphinae, buffalo, kob, warthog, waterbuck, elephant
- Vehicle type: plane
- Image information: 1297 images, each 6000x4000
- Annotation information: 10,239 boxes
- Typical animal size in pixels: 47
- License: CC-BY-NC-SA 4.0
- Code to render sample annotated image: preview-delplanque-mammals.py
- Shortcode: delplanque-mammals
SAVMAP (UAV images of Namibian wildlife)
Reinhard F, Parkan M, Produit T, Betschart S, Bacchilega B, Hauptfleisch ML, Meier P, Joost S, Tuia D. Near real-time ultrahigh-resolution imaging from unmanned aerial vehicles for sustainable land use management and biodiversity conservation in semi-arid savanna under regional and global change (SAVMAP). Zenodo.
- 3GB, downloadable from Zenodo or Hugging Face
According to the documentation provided with the Hugging Face version of the dataset:
"This dataset is not identical to the Zenodo version. The images have been cropped to 2000*2000 and only a portion of the negative samples (i.e. empty images) have been selected. The dataset comprises 3545 negative samples and 379 positive samples."
For both versions:
- Categories: animal
- Vehicle type: UAV
- Typical animal size in pixels: <100
- License: AFL-3.0
- Shortcode: reinhard-savmap
For the Zenodo version:
- Metadata in .geojson format
- Image information: 659 images, each 4000x3000
- Annotation information: ~7.5k polygons, though those really approximate boxes, and there are a smaller number of unique annotations. But O(thousands).
- Code to render sample annotated image: preview-reinhard-savmap.py
For the Hugging Face version:
- Metadata in Parquet format
- Image information: 3,924 images, each 2000x2000
- Annotation information: 1283 boxes
- Code to convert Parquet to COCO and render sample images: convert-savmap-huggingface.py
Big Bird
Wilson JP, Amano T, Bregnballe T, Corregidor-Castro A, Francis R, Gallego-García D, Hodgson JC, Jones LR, Luque-Fernández CR, Marchowski D, McEvoy J. Big Bird: A global dataset of birds in drone imagery annotated to species level. Remote Sensing in Ecology and Conservation. 2026.
23,865 images with species-level annotations for 100 species; only a subset have boxes (49,490 bird annotations on 4,824 images).
- Annotated data is ~45GB, downloadable from UQ eSpace
- The annotated subset is also available on LILA
- Metadata in labelme format
- License: "Permitted reuse with acknowledgement" (license details)
- Code to render sample annotated image: preview-wilson-bigbird.py
- Shortcode: wilson-bigbird
Datasets added recently that I haven't had a chance to dig into yet
BAMBI dataset
Praschl C, Schedl DC, Maschek A, Probst K, Stöckl A, Böss L, Jantsch W, Leitner H, Wohlfahrt S, Leiler H, Schneeberger R, Wipplinger A, Willemsen F, Bronner G, Jawecki B, Eppich G. The BAMBI Dataset: Multimodal Nadir UAV-Recordings of Forest Wildlife. Zenodo, 2026.
- ~75GB, downloadable from Zenodo
- Metadata in MOT format
- Categories: wild boar, red deer, roe deer, fallow deer, alpine ibex, chamois, bird, human, dog, hybrid pig
- Vehicle type: UAV
- Image information: 389 paired RGB+thermal images
- Annotation information: ~5100 annotated tracks with ~93k annotated keyframes, interpolated to ~1.2M boxes
- Typical animal size in pixels: TBD
- License: CC-BY 4.0
- Code to render sample annotated image: TODO
- Shortcode: praschl-bambi
Multi-species wildlife in South African savanna
Allin P, Seydou F, Frans R, Davies A, Leslie A. Evaluating machine learning models for multi-species wildlife detection and identification on remote sensed nadir imagery in South African savanna. Wildlife Biology. 2026 Feb:e01523.
- ~450GB, downloadable from Dryad
- Metadata in COCO and label studio format
- Relevant code is here
- Categories: TBD
- Vehicle type: UAV
- Image information: TBD
- Annotation information: TBD
- Typical animal size in pixels: TBD
- License: TBD (Dryad page says "public domain", but I'm not sure that's accurate)
- Code to render sample annotated image: TODO
- Shortcode: allin-nadir
BOEM Birds
Ke T, Koneff MD, Lubinski BR, Robinson L, Fronczak DL, Fara LJ, Landolt KL, White TP. Code, imagery, and annotations for training a deep learning model to detect wildlife in aerial imagery. U.S. Geological Survey data release, 2024.
- Shortcode: landolt-boembirds
Datasets available by request
Identification of free-ranging mugger crocodiles by applying deep learning methods on UAV imagery
88,000 images focusing on the mugger’s dorsal body. The data was collected from 143 individuals across 19 different locations along the western part of India.
- 1.5 GB, downloadable via http from Dryad (download link)
- Categories: mugger crocodile
- Vehicle type: drone
- Image information: 88000 RGB images
- Annotation information: individual ID
- Typical animal size in pixels: 1000
- License: CC0
Whales from Space
633 satellite image chips with boxes on whales
Cubaynes HC, Fretwell PT. Whales from space dataset, an annotated satellite image dataset of whales for training machine learning models. Scientific Data. 2022 May 27;9(1):245.
- 10 MB, downloadable via http from by request (download link)
- Metadata in csv, shapefile format
- Categories: southern right whale, humpback whale, fin whale, grey whale
- Vehicle type: satellite
- Image information: 633 RGB images (150x150 chips)
- Annotation information: boxes
- Typical animal size in pixels: 50
- License: variable
Publicly-available models for wildlife detection in drone/aerial images
This section lists ML models one can download and run locally on drone/aerial images of wildlife (or use in cloud-based systems). This section does not include models that exist in online platforms but can't be downloaded locally.
I am making a very loose effort to include last-updated dates for each of these, but I'm not digging that deep; if someone trained a detector in 2016 that is totally obsolete, but they corrected a bunch of typos in their repo in 2023, they will successfully trick my algorithm for determining the last-updated date.
- Conservation Drones Elephant Tracker (2025, paper) (YOLOv11x elephant detector, plus tracking configuration files)
- HerdNet (2022, data) (custom detector for African mammals in aerial imagery)
- DuckNet (2025, paper) (RetinaNet on ResNet50 for ducks in UAV images)
- Global model of bird detection (2021, data, paper) (RetinaNet on ResNet50 in PyTorch) (downloadable directly, but recommended use is via the DeepForest package)
- DeepForest Livestock Detection Model (2024) (single-class detector for cows, sheep, and other large mammals in agricultural settings)
- Izembek goose detector (2023, data, report) (YOLOv5, detects birds in Izembek Lagoon in Alaska, particularly brant geese, in aerial imagery)
- OWL (Overhead Wildlife Locator) (2026) (a family of models including CNN-, Swin-, and DINOv3-based variants)
- Esri Tern Detector (2025, data) (Mask-RCNN, trained w/ArcGIS Python API, distributed as an Esri dlpk file (which is a zipped .pth file))
- Esri Arctic Seal Detector (2025, data) (Faster-RCNN, trained w/ArcGIS Python API, distributed as an Esri dlpk file (which is a zipped .pth file))
- Esri Elephant Detector (2025, data) (Faster-RCNN, trained w/ArcGIS Python API, distributed as an Esri dlpk file (which is a zipped .pth file))
- Esri human detector (2025) (Faster-RCNN, trained w/ArcGIS Python API, distributed as an Esri dlpk file (which is a zipped .pth file))
- WildlifeMapper (2024) (multi-class detector for African mammals)
Platforms/systems for wildlife detection in drone/aerial images
Platforms that are specifically related to wildlife, and have evidence of being active
Scout
"Scout is an open hardware and open source software solution designed by Wild Me to support the analysis of large volumes of data obtained from aerial surveys of wildlife."

SurveyScope
"SurveyScope is a powerful web application that leverages the latest artificial intelligence (AI) to assist in the annotation of aerial-census data."

WISDAM
"WISDAM is a free, downloadable application designed for researchers (or community groups) conducting wildlife imagery surveys from either piloted aircraft or drones."

Wildlife Annotation Tool
CVAT fork used by USGS/BOEM/USFWS for bird model training. Not publication available AFAIK.

Platforms that are specifically related to wildlife, but look inactive
WildAI
Not a ton of information as of 2026.07.23, has been "in the final stages of development" since early 2025. Web page says "Simply upload your aerial images to our platform, where our powerful AI model automatically processes and analyzes the data for you, and then seamlessly review and verify the results with complete confidence and ease."
AIDE
"AIDE is two things in one: a tool for manually annotating images and a tool for training and running machine (deep) learning models. Those two things are coupled in an active learning loop: the human annotates a few images, the system trains a model, that model is used to make predictions and to select more images for the human to annotate, etc."
As of 2026.04, the last meaningful commit was ~2022.

Platforms that aren't specifically related to wildlife, but that people use for wildlife stuff
Label Studio
Label Studio (LS) is a general-purpose platform for data annotation (not just images, all kinds of data). Broadly, it comes in two flavors: LS Enterprise is a hosted (not free) service with lots of features to manage your annotation workforce; LS Community is a containerized version with largely the same front-end, but fewer annotator management and ML features. Both are flexible and templated, to the point of maybe being a little too complicated to use for wildlife survey applications out of the box, but if this community rallies around specific templates and workflows, it may be The Right Thing.
- home
- code
- blog post from Ben Weinstein about using using LS for wildlife annotation

BisQue
BisQue stands for "Bio-Image Semantic Query User Environment", and its mostly a cloud-based collaborative annotation platform for microscopy and biomedical images, so you might wonder why it's included here... at the end of the day, it's a platform for image annotation, with support for georeferenced images, and at least two people have mentioned using it for wildlife images, so, it counts.

Zooniverse
Not directly machine-learning-related, but it seems relevant in the sense that it's a good way to collect training data, and lots of the same folks who might use ML-accelerated annotation are likely to also leverage citizen scientists.
- home
- code
- "Aerial Wildlife Surveys in Africa" project
- "Penguins from Above" project
- "Drones for Ducks" project

OSS repos about wildlife detection in drone/aerial images
...that are not already listed in the publicly-available models or systems/platforms sections above.
- OpenWildlife (mmdetection variant for duck detection in UAV images)
github.com/echonax07/OpenWildlife - POLO (modified YOLOv8 that trains on point labels)
github.com/gigumay/POLO - DeepForest (tools for object detection in aerial images, esp trees and birds)
github.com/weecology/DeepForest - UAV Thermal Wildlife Detection (training detection models on the BIRDSAI dataset)
github.com/tiffanyyk/UAV-Thermal-IR-Wildlife-Object-Detection - HealthyCountryAI (training and inference for UAV wildlife detection in Australia)
github.com/microsoft/HealthyCountryAI - kabr-tools (behavioral analysis from drone videos)
github.com/Imageomics/kabr-tools