Integral Curvature Representation and Matching Algorithms for Identification of Dolphins and Whales

ElephantBook: A Semi-Automated Human-in-the-Loop System for Elephant Re-Identification

“…Recent methods of robustly differentiating between elephant images in an open-set population rely on finding and matching the contours of the ear (Figure 9), similar to many human-expert re-identification methods like SEEK. Multi-curve matching algorithms based on human-annotated contours of elephant ears were proposed by Ardovini et al [6] and Weideman et al [71]. Weideman's CurvRank algorithm was originally designed for re-identification of whale flukes and dorsal fins.…”

“…CurvRank was initially developed to recognize individual cetaceans based on contours of flukes and dorsal fins [71].…”

NDD20: A large-scale few-shot dolphin dataset for coarse and fine-grained categorisation

“…This may seem like an almost impossible task, however it has been achieved by cetacean researchers for over forty years (21), manually identifying individuals based on prominent markings on their fins as they breach the waterline. As such it is believed that an automated photo-id process is achievable using computer vision (19), with work already beginning in this area (3,5,6,18). Development of automatic photo-id is currently limited to niche research groups who have the domain expertise in cetacean research as well as the technical and mathematical ability to develop automated systems to utilise this.…”