Mask R-CNN Based Laptop Parts Detector for Disassembly

The efficient recycling of waste electronic 3C products depends on the development of automatic disassembly process, including the robust vision detection system. Here we choose laptop parts as the target object, and build a dataset that contains 620 images. Our work focused on parts detection at disassembly scenario and implemented a universal and extensible laptop parts detector that can be applied to disassembly pipeline which based on a light and improved Mask R-CNN. After generating the mask images of parts, a rotary rectangle fitting method for parts was performed to predict the parts’ rotation angle, and apply specific rules based on quantity and space constraints to improve the accuracy.