Performance Analysis of Object Detection Algorithms for Waste Segregation

The amount of waste generated in urban cities is increasing at an alarming rate and its effective and efficient management has become nearly impossible. To combat this issue, this study has compared different object detection algorithms applied on data containing trash with no cluttered background, so as to first evaluate the different algorithms only on the basis of their inherent performance. The algorithms used for training are YOLOv5, EfficientDet and Faster RCNN. Using mAP(Mean Average Precision) as the evaluating metric, this study has found that YOLOv5 with mAP of 0.92 has outperformed EfficientDet and Faster RCNN, whose mAP was 0.846 and 0.836 respectively.