Deep Learning Models for the Identification of Coffee-Bean

Prior to roasting the beans, finding flaws in the coffee beans is the most critical step. When it comes to flaw detection, the specialist coffee bean sector still uses manual screening and sorting procedures that are labor-intensive and not standardized. This study developed a deep learning approach to identify and fix coffee bean defects. The results show that when the pooling layer was used to improve features and reduce neural complexity, some of the coffee-been features were either removed or mistakenly identified. Therefore, in order to improve feature extraction even more, a novel reduction of dimensionality technique was employed. Problems with padding and neuronal death that prevented features from propagating and blurred image borders, respectively, are both addressed by the new approach. We used images of eight different coffee bean kinds to train and evaluate the proposed detection model. There was prior evidence that the suggested strategy could reduce bias when sorting defective coffee beans. The detection accuracy rate of the suggested model was 94%. The proposed method is thus very effective in detecting coffee bean flaws across eight distinct coffee bean kinds.