Animal behavior analysis at scale
Research Methods Video data provides insights into the social dynamics of wild animals, but processing large video datasets can be expensive and labor intensive. Using a deep convolutional neural network, Schofield et al. developed an automated pipeline to detect, track, and recognize individual animals from video in the wild. They demonstrate and evaluate this approach using a multiyear chimpanzee dataset. This system allowed researchers to monitor individual animals and automatically extract behavioral and social dynamics of individuals and groups directly from video. Sci. Adv. 10.1126/sciadv.aaw0736 (2019).