A Novel Approach of Object Detection using Deep Learning for Animal Safety

Abstract

This paper uses convolutional neural networks (CNN) algorithm to detect animals. Four different Object Detection frameworks are implemented and compared their performance with our novel approach. Many algorithms of convolutional neural networks like RCNN, Faster RCNN, Single Shot detection, YOLOv3 are used in object detection tasks are compared. A customized Faster RCNN is designed with anchor manipulations This framework is found to be best suited technique for Animal Detection System giving the best time-accuracy trade-off. Experimental results shows that this method has reduced the best inference time significantly.

Topics & Concepts