Computer vision based waste classification using deep learning

Abstract

Waste management systems and their inherent problems are still a matter of great concern even amid this cutting edge of science and technologies. The root cause of this problem points to one fact - which is too much manual labor in the garbage collection, separation and recycling process - can't keep up to the pace with which garbage generation happens. In this research, We will propose a novel Deep Learning based approach of automatic separation of five kinds of waste materials namely - Kitchen Waste, Glass Waste, Metal Waste, Paper Waste, Plastic Waste, from the garbage dump for an efficient recycling process, which not only improves the efficiency of the current manual approach but also provides a scalable solution to the problem. The contributions of this project includes a fully human labelled data set consists of 2000 images of garbage dump and a real time garbage localization and classification framework based on a single stage object detection algorithm. For the baseline, we have used YOLOv4 Object Detection Algorithm and with some fine tuning, we proposed a modified object detection framework which yields a mAP of 66.08% with an inference speed of 70 milliseconds on both images and videos.