Skin diseases impact over 900 million individuals globally, presenting significant healthcare challenges, especially in underserved regions with limited access to dermatological services. This paper aims to develop Convolutional Neural Networks (CNNs) for the early diagnosis and prediction of common skin diseases, focusing on melanoma, psoriasis, dermatitis, and herpes. The primary goal is to provide an accessible tool to improve public health outcomes in areas with constrained medical services. The methodology involves compiling a comprehensive dataset of various images of diseased skin, covering different conditions. Steps include image preprocessing, augmentation, and feature extraction using CNN models, leading to disease classification. Our approach enhances the proficiency and accuracy of skin disease diagnosis, proving particularly useful where access to medical professionals is restricted. This research demonstrates that fine-tuning CNN architectures with the Adam optimizer significantly improves the accuracy of skin disease detection, offering a valuable tool for real-time diagnosis and severity assessment of skin conditions.