Skin diseases affect millions worldwide, yet accurate and timely diagnosis remains challenging due to the diversity of conditions and class imbalances in available datasets. This project, Channelling Fairness: Class Imbalance-Aware Skin Disease Recognition via Fair Channel Enhancement Module, proposes a deep learning framework that enhances classification performance while addressing skewed data distributions. The system integrates the Fair Channel Enhancement Module (FCEM) with ResNet50 and SE-EfficientNet-B3 architectures to extract discriminative features from skin images effectively. FCEM dynamically adjusts feature channels to improve representation of underrepresented classes, reducing bias in predictions. The web-based application, developed using HTML, CSS, JS, and Python with Flask, includes modules for user registration, login, and skin disease prediction. The model is trained on a publicly available skin disease image dataset from Kaggle, encompassing multiple disease categories. Experimental evaluations demonstrate improved accuracy, balanced sensitivity across classes, and robustness in classifying both common and rare skin conditions. This approach provides an accessible platform for researchers and practitioners to study skin disease recognition and supports the development of AI-assisted diagnostic tools. The results indicate that addressing class imbalance through channel enhancement significantly enhances model fairness and predictive reliability.

Keywords: Skin disease, Class imbalance, Deep learning, ResNet50, EfficientNet-B3, Fair Channel Enhancement, Image classification, Flask, Bias mitigation, Medical imaging