Skin Cancer Detection and Classification

Skin cancer is a serious health issue, and early detection significantly improves treatment outcomes. Our study proposes a skin cancer detection system using a transfer learning model, which leverages pre-trained models to enhance classifi-cation accuracy. This is particularly useful when working with limited data, such as our ISIC dataset of 2,500 images across five skin cancer diseases such as squamous cell carcinoma, actinic keratoses, vascular lesions, pigmented benign keratosis, and melanoma. Transfer learning offers key advantages, including reduced training time and computational resources, as the model already understands basic image features. Our model achieves 90% accuracy, demonstrating its potential in aiding early diagnosis. However, it has limitations, such as struggles with rare or less represented diseases and the potential risk of misclassifying aggressive cancers like melanoma. In conclusion, while the method shows promise, further im-provements are necessary. Expanding the dataset and refining the model could enhance its accuracy and reliability for broader clinical use.