Quality Assessment of Coffee Beans Using Convolutional Neural Networks with Wavelet and CLAHE Techniques

This paper presents an analytical study comparing different filtering techniques applied to a Convolutional Neural Network (CNN) for coffee bean classification. The results demonstrated that the CLAHE (Contrast Limited Adaptive Histogram Equalization) filter achieved the highest performance, with an accuracy of 0.8875 on the test set. The findings indicate that applying filtering techniques can enhance the performance of the ResNet18 network. CLAHE’s effectiveness is attributed to its ability to improve image details and contrast, leading to superior classification results. This study underscores the potential of advanced filtering methods to boost CNN performance in image classification tasks.