Plastic Waste Detection on Rivers Using YOLOv5 Algorithm
Gilroy Aldric Sio, Dunhill Guantero, Jocelyn F. Villaverde
Abstract
Building sustainable, clean communities have always been a challenge, especially with the surge in population that increases waste and rubbish production. A higher pollution level results from increased rubbish production, which has a variety of negative repercussions on the neighborhood. In light of this, the study is focused on detecting plastic waste and garbage on rivers through the creation of a new system with the utilization and application of the YOLOv5 algorithm. The researchers used a Raspberry Pi Model 4 B as a microcontroller for the design and implemented a 5MP Camera Module and a USB camera to acquire images of floating plastic bottles on the river. The training procedure of the algorithm is carried out initially through the creation of a custom dataset and is processed on a computer. Based on the measured metrics and evaluated confusion matrix, the model produced an overall accuracy of 84.298% in detecting plastic bottles on the river. In addition, the model also yielded a precision rate of 79.14% and a recall rate of 57.37%, which indicated a considerable quality for object detection.