Automated Animal Detection and Classification using Machine Learning Approach

In this paper, machine learning approach algorithm models such as the Regionbased Convolutional Neural Network (R-CNN) classification model, Single Shot multi-box Detection (SSD), and You Only Look Once (YOLO) regression model are used for the detection and classification of animals. For experimentation purposes, 2000 datasets of four classes of animals (elephant, cheetah, cow, and horse) were collected using remote surveillance cameras. By eliminating bounding boxes and gradually reducing convolution filters, SSD and YOLO improve computer vision tasks in speed, accuracy, and efficacy for relatively low-resolution images compared to R-CNN. The proposed classification algorithm model R-CNN gives a mean average precision (mAP) of 95.22%, whereas the regression algorithm models SSD and YOLO give mAPs of 98.32% and 98.89%, respectively.