Lightweight Deep Convolution Neural Network for Green Coffee Bean Defects Detection

The demand for high-quality coffee in the global market has increased. However, large labor resources are still required for sorting beans in coffee production due to different manual screening standards. Thus, maintaining consistency in the quality of coffee beans poses a challenge. However, computer vision (CV) based on lightweight deep convolutional neural networks (DCNNs) was used to detect coffee bean defects and evaluate the proposed model’s structure. The proposed method was trained with the public coffee bean database. The experiment result showed a high recognition performance even with a small number of parameters achieving 96.00% of the F1-Score.