{"version":1,"kind":"Article","sha256":"","slug":"156","location":"","dependencies":[],"frontmatter":{"title":"Identifying Individual Animals using Ranking, Verification, and Connectivity","abstract":"In this thesis we address the problem of identifying individual animals using images in\nthe context of assisting an ecologist in performing a population census. We are motivated\nby events like the “Great Zebra Count” where thousands of images of zebras and giraffes\nwere collected in Nairobi National Park over two days. By grouping all images that con-\ntain the same individual we can census these populations. This problem is challenging\nbecause images are collected outdoors and contain occlusion, lighting, and quality varia-\ntions and because the animals exhibit viewpoint and pose variations.\nOur first contribution is an algorithm that ranks a database of images by their sim-\nilarity to a query. A manual reviewer inspects only the top few results for each query —\nsignificantly reducing the search space — and determines if the animals match. Using\nthis algorithm alone, we analyzed the images from the Great Zebra Count and performed\na population census. Our second contribution is a verification algorithm that determines\nthe probability that two images are from the same animal, that they are not, or that there is\nnot enough to decide. This algorithm is used with the ranking algorithm to re-rank results\nand automatically verify high confidence image pairs.\nOur third contribution is a semi-automatic graph identification algorithm. The ap-\nproach represents each image as a node in the graph and incrementally forms edges be-\ntween nodes determined to the same animal. The ranking and verification algorithms are\nused to search for candidate edges and estimate their probability of matching. Based on\nthese probabilities, edges are prioritized for review and placed in the graph when they\nare automatically verified or manually reviewed. Redundant connections are added to\ndetect and recover from errors. A termination criterion determines when identification is\nfinished. Using the graph algorithm we perform a population census on the scale of the\nGreat Zebra Count using less than 25% of the manual reviews required by the original\nmethod.","license":"CC BY","keywords":[],"authors":[{"name":"Jonathan Crall","orcid":"0009-0008-8455-7514","affiliations":[],"corresponding":true}],"date_submitted":"2024-01-31T21:05:48.000Z","external_publication_id":156,"revision_cids":["bafkreibhuuogniwgsozq6zqgpfg622ktaz4e4xtspvxv6be7zoaknlp7oi","bafkreicmyxjhueivvlucnysc5sgmdnxgsmn5wryaodmdwl7xhku5o4dwki"]},"mdast":{"type":"root"},"references":{"cite":{"order":[]},"data":{}}}