Coffee Bean Diseases Classification Using Convolutional Neural Network Model

The most important stage before bean roasting is identifying defects in coffee beans. The specialized coffee bean industry continues to employ non-standardized, labor-intensive manual screening and sorting methods to perform fault detection. This work created a deep learning method that recognizes faults in coffee beans and tackles these issues. The findings reveal that some of the coffee-been features were eliminated or incorrectly recognized when the pooling layer was applied to improve features and decrease neural complexity. As a result, a unique reduction of dimensionality technique was used to further enhance feature extraction. The new model also surpasses the issues with padding that gave rise to blurred image boundaries and neurons dying that inhibited feature propagation. The suggested detection model was trained and tested using pictures of eight distinct varieties of coffee beans. The proposed method's ability to lessen bias when sorting flaws in coffee beans was previously demonstrated. The proposed model's detection accuracy rate was 98%. Thus, in the categorization of eight different types of coffee beans, the suggested approach is quite accurate in identifying coffee bean defects.