It’s important to catch skin cancer early, as it makes all the difference in successful treatment. To help with this challenge, we developed an AI system that analyzes images of skin lesions to determine if they are benign or malignant. Unlike some previous studies, we trained our model on a dataset that includes people of all ages, making it more applicable to real patients. The key to our system’s performance is how we prepare the images. We resize them, adjust brightness, and use normalization. A specially designed U-Net model then acts as a digital highlighter, precisely outlining the lesion and separating it from the surrounding skin. This allows the subsequent EfficientNet AI to focus only on essential features. Our results are encouraging, with the system achieving 78% accuracy and demonstrating high stability. This proves that intelligent image preparation is just as critical as the AI model itself for reliable diagnosis.