NESTLER Wild Animal Recognition Video Dataset

This video dataset was created under NESTLER (oNe hEalth SusTainabiLity partnership between EU-AFRICA for food sEcuRity) Horizon Research Project, Funded by the European Union under Grant Agreement no.101060762.

-------------------------------------------------------------------------------------------------------------------------------------------

The folder "Foxes-Vultures-Ravens-Jackals_Classification" contains videos of four animal classes, namely Foxes, Jackals, Ravens and Vultures. The videos were captured during day and night using RGB and InfraRed sensor cameras.  

It can be useful in identifying one of these four animals or it can be used along with other data utilizing each class as part of wider animal categories.

The videos were captured at RAKOVO, Sliven region, Bulgaria, on 14-19 September 2024 by RINISOFT. 

For video capturing Foxes, Jackals and Ravens a HD Camera Balever BL480LP 4G was used and specifically for Foxes and Jackals videos were captured with a Night Vision mode. The video resolution for these categories was 1920x1080.

For Vultures there was used a Digital Camera, Casio EX-H15 and a Fujifilm X-T4 with a video resolution set at 720p.

--------------------------------------------------------------------------------------------------------------------------------------------

The dataset also contains videos of 12 classes of wild animals that can be found in Africa. Namely, the 12 classes are  Baboons, Buffaloes, Elephant, Girafes, Gorillas, Hippopotamus, Impala, Lions, Rhinocerus, Topi, Warthog, Zebra.

The importance of the dataset is obvious; it is a rare dataset since it is not easy to capture these wild animals at their natural environment. It may be dangerous to approach and visit these animals and also the road facilities are bad.

The videos were captured with RGB cameras at Rwanda from RAB and Uganda from CTPH. 

--------------------------------------------------------------------------------------------------------------------------------------------

The folder "Elephants-Gorillas-Kobs_Detection_Dataset" contains the extracted frames which are accompanied by their respective bounding box annotations making this part of the dataset suitable for detection tasks.

The videos were captured in Uganda at different periods of 2025 (January, February, June) using a HD camera (1920x1080) from CTPH.

-------------------------------------------------------------------------------------------------------------------------------------------- 

Safari_Wild_Animal_Detection.zip:

This dataset contains annotated wildlife images collected in three national parks in Uganda (Kibale National Park, Queen Elizabeth National Park, and Bwindi Impenetrable National Park) between February and June 2025. It includes key species such as elephants, chimpanzees, mountain gorillas, hippopotamuses, Uganda kob, and warthogs. All images are annotated with bounding boxes and class labels in COCO/YOLO-compatible formats. The dataset includes 9 videos, 1,173 frames, and 3,909 bounding boxes. It is intended to support the development and benchmarking of wildlife detection and classification models, as well as research on AI for environmental monitoring and sustainability.

The images were captured using a Canon EOS M50 MK II and an iPhone 12 Pro Max (12MP Ultra-Wide, Wide, Telephoto, Night, HDR, ProRAW). The video resolution is 1920x1080. Annotation was performed using bounding boxes and class labels, with outputs provided in COCO/YOLO-compatible JSON and TXT formats.

All data were collected in line with ethical wildlife observation practices. No animals were disturbed, trapped, or manipulated, and camera traps were deployed in accordance with park regulations to minimise disturbance. All necessary permits for photography and data collection were obtained from relevant authorities. The dataset contains no personal data or identifiable human subjects, ensuring full compliance with EU data protection (GDPR) requirements.