Research overview: deep-learning re-identification of individual cats

Emerging deep-learning methods aim to automate not just species classification but individual identification, re-identifying specific cats from their coat patterns to build capture histories for capture-recapture. This overview surveys how convolutional and metric-learning approaches learn embeddings that place images of the same individual close together, enabling automated matching of tigers, leopards, jaguars, and other patterned felids across photographs. It discusses the promise of reducing the labor-intensive manual matching that underpins density estimation, the challenges of pose, lighting, and partial views, the need for confirmed-identity training data, and the importance of human verification to avoid false matches that corrupt abundance estimates. The overview positions automated re-identification as a frontier that could greatly scale rigorous density monitoring, while cautioning that, like species classification, it requires careful validation before trusting its matches in scientific analysis.