YOLOv8 Transfer Learning In Smart Bin Garbage Sorting Machine

This paper proposes an idea to efficiently utilize waste products by integrating hardware and software for garbage sorting. The Smart Bin system aims to efficiently categorize and sort waste using advanced image processing and deep learning techniques. A metal sensor is used to identify metal items, while image processing is used to identify other items. In the image processing stage, images are enhanced with kernel filters before using YOLOv8 to predict labels and identify the location of garbage in the images. During the training process, the first 10 layers are frozen and then retrained using a custom image dataset supplemented with images captured from the machine itself. Edge detection in preprocessing further enhances garbage edge predictions and separates objects on the conveyor belt in the sorting machine. The results from the experiment demonstrate the system's capability to identify both the location and type of garbage. This concept is proposed to help solve pollution and waste problems and to inspire the further development of other new projects in the future.