In this thesis we address the problem of identifying individual animals using images in the context of assisting an ecologist in performing a population census. We are motivated by events like the "Great Zebra Count" where thousands of images of zebras and giraffes were collected in Nairobi National Park over two days. By grouping all images that contain the same individual we can census these populations. This problem is challenging because images are collected outdoors and contain occlusion, lighting, and quality variations and because the animals exhibit viewpoint and pose variations.

Our first contribution is an algorithm that ranks a database of images by their similarity to a query. A manual reviewer inspects only the top few results for each query --- significantly reducing the search space --- and determines if the animals match. Using this algorithm alone, we analyzed the images from the Great Zebra Count and performed a population census. Our second contribution is a verification algorithm that determines the probability that two images are from the same animal, that they are not, or that there is not enough to decide. This algorithm is used with the ranking algorithm to re-rank results and automatically verify high confidence image pairs.

Our third contribution is a semi-automatic graph identification algorithm. The approach represents each image as a node in the graph and incrementally forms edges between nodes determined to the same animal. The ranking and verification algorithms are used to search for candidate edges and estimate their probability of matching. Based on these probabilities, edges are prioritized for review and placed in the graph when they are automatically verified or manually reviewed. Redundant connections are added to detect and recover from errors. A termination criterion determines when identification is finished. Using the graph algorithm we perform a population census on the scale of the Great Zebra Count using less than 25% of the manual reviews required by the original method.

The bibtex citation entry is:

@phdthesis{crall_identifying_2017,
  address = {Troy, NY},
  author = {Crall, Jonathan P.},
  school = {Department of Computer Science, Rensselaer Polytechnic Institute},
  title = {Identifying {Individual} {Animals} using {Ranking}, {Verification}, and {Connectivity}},
  type = {Ph.{D}. {Thesis}},
  year = {2017}
}

The full text PDF is available on GitHub to be viewed or downloaded.

Google slides hosts my original defense presentation (2017-07) and candidacy presentation (2016-05). These can also be found on IPFS via the following CIDS: QmezNaQ2GypcN8951DxXWEkZdwxtZb8tohWKjHqqGVRDUi and QmZDsmeJTwiQKFeCAWA5Vpq1hrK96xkmmLfgkSQ2DDp5qs.

The final state of the repo including all images needed to reproduce the PDF have been uploaded to IPFS (as of 2022-04-09). The root CID is: QmTD1nZ4pbrB1SnjkLGt9Cs37mZbabXqjn6YZaAKVEoSvY. And the main relevant data CIDS are:

QmUkJhNSLDSgBnvxu5TLiXGU8LCR7tfS9s2meWCZZQ6bnL crall-thesis-2017/guts
QmScScfpeKA2CtShuin5o5af5MN3NFhQBLeDqEUb435qXu crall-thesis-2017/notes
QmPmpgBGw2kHPRVBg6z9m28PNLAzCKMe29QvJ3YvWZRo1G crall-thesis-2017/tmp/figures1
QmUYQFvdWEVYEsmf5wNfHTPzWrHtaiZjcSanf5FsD74NYu crall-thesis-2017/tmp/figures2
QmNqvbG1ng6z3bATR4XRyPiSGVExHaoLUQszqYFeARrBoc crall-thesis-2017/tmp/figures3
QmeuC9tYezC75yRLrG7tvCT7tXrQYsLWtvVw62ZYGDEBBw crall-thesis-2017/tmp/figures4
QmWmkVJv48A5RTGSm8H92qUHvcKeKsKugHbYsua8ktMjQV crall-thesis-2017/tmp/figures5
QmTs6PB5148DouDWKAv8WB9Q2zdx3crCBb3W88bHjLLkfM crall-thesis-2017/tmp/figuresC
QmUf1YkuEdmEczm6j237rgtUwSKaNkL4uCCq4QFZrfDL2c crall-thesis-2017/tmp/figuresGraph
QmajtiHvGuQ4vbVPH36R6E5KAxpQao5R7XGre8aqa8iWPK crall-thesis-2017/tmp/figuresX
QmXwq83V6WBtd8vSV3JNHTyYvH3HiU6CzA46jDtSV6xGwD crall-thesis-2017/tmp/figuresY
QmZKeYDeVkKMbmEvL7ue8r8uG2TqB6aJXhX2dhNLKsawD3 crall-thesis-2017/tmp/figures_graph
QmZuc9FyH6mJ1hsMwKv77rbyGeWzLGCjrCd8gyXNtCh9JJ crall-thesis-2017/tmp/figures_new3
QmQd9zEb2SxbweNcLMGmnx34AvBJWCp4PUbrdgZkP1RTVG crall-thesis-2017/tmp/figures_pairclf
QmTD1nZ4pbrB1SnjkLGt9Cs37mZbabXqjn6YZaAKVEoSvY crall-thesis-2017

The software developed while working on this thesis is free and open source.

The primary software is the IBEIS package: https://github.com/Erotemic/ibeis

An older version of the software is the HotSpotter application: https://github.com/Erotemic/hotspotter

The modern versions of these algorithms are brought together in the WildBook framework for mark-recapture, molecular ecology, and social ecology studies: https://github.com/WildMeOrg/Wildbook

Services providing a web-based infastructure for the algorithms described in this thesis and others are provided by the WildMe organization: https://www.wildme.org/

Parham, Jason. "Animal Detection for Photographic Censusing" RPI PhD Thesis 2021

Weideman, Hendrik. "Contour-Based Instance Recognition of Animals" RPI PhD Thesis 2019

Crall, Stewart, Bertger-Wolf, Rubenstein, and Sundaresan. "HotSpotter - Patterned Species Instance Recognition" WACV 2013

Parham, Crall, Stewart, Berger-Wolf, and Rubenstein, "Animal population censusing at scale with citizen science and photographic identification" AAAI 2017

The following is a selection of figures from the this that provides a visual summary of the contents.