Multi-Animal Recognition in Zoo Cluster using Deep Learning Technique

Abstract

The accelerated development of new methodological approaches and techniques for quantifying real-time animal recognition can be credited to the concern with the question of how the brain originates and organizes animal recognition. Even while the latest development in deep learning and computer vision have made it possible to estimate a single animal's pose, expanding this capability to several animals creates a new set of challenges for researchers who study animals in their natural environments. In this article, we will discuss the Radial Basis Function Networks (RBFNs) technique, which is a deep learning system designed for the recognition of several animals. This system makes it possible to execute a broad range of workflows for the labeling of data, the training of models, and the generation of conclusions based on data that has been seen executed several times before. RBFNs are preloaded with a graphical user interface that is easily accessible, a uniform data model, and a reproducible setup system. We deployed RBFNs to datasets spanning a wide range of zoo animals to thoroughly analyze each strategy and architecture, and we compared it to other current approaches. RBFNs can achieve both improved accuracy and increased speed. This enables the utilization of RBFNs in real-time applications, such as those in which we demonstrate the face pose of one animal based on the tracking and detection of another animal.