WildLive

Read full text
Authors: Dat Nguyen Ngoc, Tom S Richardson, Matt Watson, Kilian Meier, Jenna Kline, Sid Reid, Guy Maalouf, Duncan Hine, Majid Mirmehdi, Tilo Burghardt
Cited by: 0

Abstract

WildLive dataset contains 215,800 bounding boxes and animal segmentation masks of 291 zebra, giraffe, and elephant tracklets, plus 84 point tracks across 22 UAV-acquired 4K video sequences. This totals 19,139 frames recorded on site at the Ol Pejeta Conservancy in Kenya. The data was acquired using DJI Mavic 3 Enterprise and Pro drones plus a custom-built quadcopter for wildlife missions. Overall, the dataset provides animal bounding boxes and SAM2 segmentations, tracklet IDs, as well as manually corrected, sparse LK pixel trajectories. The accompanying publication can be found at: https://research-information.bris.ac.uk/en/publications/wildlive-near-real-time-visual-wildlife-tracking-onboard-uavs/ The accompanying project website can be found at: https://dat-nguyenvn.github.io/WildLive/

Fields of Study & Subjects

Video Surveillance and Tracking Methods Advanced Neural Network Applications Wildlife Ecology and Conservation

Publication details

Journal / source
Bristol Research (University of Bristol)
Publication date
1 January 2026
Language
EN
Document type
Dataset