Deep Learning Based Multiple Object Detection

Abstract

Improvement of automation increases rapidly by deep learning-based object detection. If the small objects are detected by YOLO V3 network, the identification result will not be much successful, so we introduce new technology to find the very small objects by 8X downsampling rate. We modified the YOLO V3 network using LaSOT with the help of anchor values and prediction layer of 4X downsampling layer, this helps to detect very small objects very accurate.