Multi Detection and Segmentation Coconut Shell for Charcoal Briquette Using Mask R-CNN

Coconut shell are the raw material for making high-quality charcoal briquettes, and their quality dramatically affects the quality of the final product. Therefore, assessing and determining the quality of coconut shells is very important in production. This research uses the Mask R-CNN instance segmentation algorithm with ResNet 101 to detect and classify coconut shell quality as a raw material for charcoal briquettes. Our model consists of four object classes: clean and dry coconut shell, fibrous coconut shell, wet coconut shell, and both wet and fibrous coconut shell. Researchers collected a dataset consisting of 651 images with a total of 8922 annotated objects directly from various sources of waste: traditional markets, coconut entrepreneurs, and household waste, which were then divided into 520 images for training and 131 images for validation. The research shows that the model achieves a mean precision value (mAP) of 0.98 on the training data and 0.94 for the validation data, indicating good prediction accuracy. Overall, the Mask RCNN model trained in this research shows excellent potential for use in producing high-quality charcoal briquettes.