Computer Vision for Animal Behavior Analysis

Knowledge of animal behavior is paramount to develop studies not only in the field of ecology, ethology, veterinary, and animal welfare. Conventional behavior observation techniques are highly dependent on manual observation which is cumbersome, subjective and can carry out its own errors as done by human beings. Computer vision has recently become an important unsolicited, scalable, non-invasive method to analyse animal behavior. The following paper gives an overview of methods of computer vision that were used in the study of animal behavior, namely object detection, pose estimation, tracking and activity recognition. It talks about the application of deep learning algorithms and data fitting of diverse organisms and environments, ranging between laboratories and wild. Computer vision incorporated into this field can increase accuracy and raise the opportunity of real-time surveillance, early detection of the disease, and behavioral phenotyping.