Deep Learning Based Approaches for Animal Pose Estimation (APE): A Survey
Syed Adnan Ali Zaidi, Asim Imdad Wagan · 2024
Animal pose estimation aims to detect the body parts and joints of animal to form a basic skeleton, on the basis of data collected in the form of images and annotations. Recently interest in animal pose estimation has increased after the success of research in Human Pose estimation, the main reasons being scientific investigation in the form of noninvasive behavioral tracking. In this paper, we present a survey on current deep learning methodologies animal pose estimation in single image or sequence of images. Although, the recently developed deep learning-based solutions have achieved high performance in human pose estimation, there still remain challenges for animal pose estimation due to a different body structure, lack of training data, and many other reasons. This work attempts to create a taxonomy that summarizes important aspects of deep learning for approaching the problem of Animal Pose with new vigor. We shall review the details of the proposed architectures, the main datasets, approaches, backbone architecture and measurement metrics. It is proposed to discuss and summarize the main methods proposed so far with particular interest on how deep learning models are used in animal pose estimation, discussing their main features and identify opportunities and challenges for future research.