Application development for plastic bottle detection using deep learning

Fong, Yun Xin (2024) Application development for plastic bottle detection using deep learning. Final Year Project, UTAR.

Preview

PDF Download (3120Kb) | Preview

Abstract

Nowadays, recycling centers still rely on human workers which is low efficiency and working environment is bad for the human workers. Hence, deep learning is introduced to detect the plastic bottles on the moving conveyer belt in the recycling centers. In this project, three pre-trained deep learning models is selected to train and detect the plastic bottles. The three selected pre-trained deep learning models are YOLOv8, Faster R-CNN and SSD. The results show that YOLOv8 achieved the highest mean average precision for the custom dataset which is 0.923 compared to Faster RCNN and SSD. Thus, YOLOv8 is selected and further tested with the real video from the recycling center to detect the plastic bottles on the conveyer belt. In the video, YOLOv8 achieved an average precision of 0.3026 in detecting the plastic bottles, but the average precision significantly improved to 0.6783 when the waste products is less overlapping on the moving conveyer belt. The application had passed the user satisfactory survey and user acceptance test, so it is easy to be used for people who does not have knowledge in deep learning.

Actions (login required)