Recognition of Plastic Bottle Using Improved U-Net
Daiki Ideta, Tohru Kamiya · 2022 22nd International Conference on Control, Automation and Systems (ICCAS) · 2022
This paper focuses on separation rubbish, which is one of the causes of global warming. Currently, recyclable waste including garbage is sent to the waste disposal site in an unsorted state, and it is separated by human source. This is difficult work and a cause of manpower shortage. To overcome those problem, it is necessary to introduce automatic sorting algorithms on separation rubbish process. In this study, we focused on separation rubbish process especially plastic bottles based on image recognition technique using a deep learning approach. We implement an improved U-Net as the deep learning scheme to increase segmentation accuracy. As a result, we obtained the segmentation accuracy with 0.789 of cap part and 0.972 of body part on plastic bottle respectively.