Exit Menu

A deep learning model for global camera trap labelling


Project description

Conservationists are increasingly using mobile “camera-traps” to monitor animal activity around the world, generating vast amounts of images and videos. After a field study has been conducted, each image / video is manually processed to identify what species are found. This process takes a lot of manual effort, some of which has been put to volunteers in citizen science projects such as Zooniverse. Work has begun on automating the labeling process with promising results and this project aims to use Artificial Intelligence to recognise species, integrate other data (including satellite images), and build spatial models of species to demonstrate how camera-traps can be used to better understand  animal behaviour over large geographical regions.