Reliable Multiple Object Detection on Noisy Images by Using Yolov3

Abstract

Object detection achieved very good performance by using deep learning models but there is a problem with noisy images. Due to the presence of noise in images, it is difficult to detect the object accurately. The main objective is to detect multiple objects in noisy images by using YOLOV3 approach. Generally, the CNN and R-CNN family algorithms extract the feature maps by using convolution operation with the striding method and objects detected by using neural network. But YOLOV3 algorithm directly applied on entire image and predict the bounding boxes along with labels and scores. In this article, the input noisy images smoothened by using median filter then YOLOV3 performs detection operation on entire image. Hence, YOLOV3 detects the object faster as compared with the other deep learning algorithms.