Coffee Bean Detection Using Mask R-CNN

Coffee Beans are a commodity that is usually used by coffee businesses to process coffee. Currently, the detection of coffee beans relies primarily on manual and visual inspection which takes time and may be inefficient. However, mechanical sorting machines can be used but it only sorts the beans based on their weight and size, and it is likely to damage the beans because of its procedure. To detect a coffee bean from foreign objects, the researchers will use an instance segmentation called Mask R- CNN. This study aims to create a system using Raspberry Pi 4 to create a detection and classification system that can detect coffee beans from other items that are surrounded by it. The researchers used foreign objects such as monggo seeds, small rocks, coins, and batteries to test the system. Having a total of 20 trials for the data gathering, the system successfully detected a total of 18 out of 20 coffee beans, achieving an accuracy of 90%.