| Title |
Fig. 3. Use of computer vision algorithms to analyse UAV footage. A - simultaneous automatic detection and tracking of two ungulate species with visualised bounding box objects (bbox) for two video frames using trained models (zebras - blue, impala - white). Example of automatic detection of animals on a complex background (B) using the Sobel operator function to detect boundaries and (C) rendering them onto the original image. D - A summary infographic demonstrating the current capabilities of using computer vision to: recognise the species identity of mammals by their position in space (purple), track their movements (pink), individually identify individuals (green), categorise body poses (orange), and reconstruct habitats, including damaged ones (blue). Illustrations are under an open licence for reuse: A - Koger et al., 2023; B, C - Lee et al., 2021; D - Tuia et al., 2022 |