A Wavelet Enhanced Deep Learning Model for Mineral Image Classification: WFeedNet

This project focuses on developing a wavelet-enhanced deep learning model, named WFeedNet, for the classification of mineral images. The model integrates wavelet transforms with deep learning architectures, such as Wave-TFNet, Wave-GCN, and Wave-SARNet, to efficiently process and classify images into mineral types. Wavelet decomposition is applied to the images to extract multi-scale features that enhance the model's ability to identify fine-grained patterns in the images. The deep learning components leverage these features to learn complex relationships within the mineral images. Additionally, Grad-CAM is implemented to visualize the areas of the image that influence the model's decision-making process, providing interpretability to the classification. The proposed model is trained on a dataset containing images of minerals, including Biotite, Bornite, Chrysocolla, and others. The model's performance is evaluated using metrics such as accuracy, precision, recall, and F1-score, demonstrating its capability to achieve high accuracy while offering insights into the decision-making process. This study contributes to improving the classification accuracy of mineral images and offers a more transparent approach to deep learning-based classification tasks.

Keywords: Wavelet Transform, Deep Learning, Mineral Classification, Wave-TFNet, Wave-GCN, Wave-SARNet, Grad-CAM, Image Processing, Feature Extraction, Interpretability.