Comparison of YOLO and Faster R‐CNN on Garbage Detection

Abstract

The huge garbage management is becoming a challenging task in India. Major parts of them are plastics, which are found in beaches and seashores. Proper treatment and recycling of such plastics is very crucial to compensate for the ever-growing demands of waste management. Garbage collection is the first step toward it. Traditional manual garbage collection is a labor-intensive process and is unable to match the ever-growing demands of city garbage. For this detection of plastics, an object detection technique has been developed. The algorithm is developed in tensor flow to make use of transfer learning. A comparison is made between the two major object detection techniques YOLO and Faster R-CNN. Faster R-CNN though comparatively slower, is found to be more accurate, hence is chosen to fit the requirement. This project further aims to develop an autonomous robot, which collects the garbage without any human intervention and increase the efficiency of garbage collection.