Vision-Based Biometrics for Conservation

Sai Ravela, James Duyck, Chelsea Finn

Abstract: Identifying individuals in photographs of animals collected over time is a non-invasive approach that enables ecological studies and conservation planning. Here we propose SLOOP, the first image retrieval system incorporating interactive image processing and matching tools with relevance feedback from crowdsourcing to solve large-scale individual identification for multiple species. One outcome is an advance in matching and image retrieval methodology; another is the creation of a community-based individual identification system that enables conservation planning.