A Comparative Study of Lightweight Convolutional Neural Networks for Skin Cancer Classification

An early and accurate diagnosis of skin cancer is critical for effective treatment and improved patient outcomes. In this paper, we propose and evaluate two lightweight convolutional neural network (CNN) architectures for skin lesion classification using the PH۲ dataset. The first model is a simple CNN with three convolutional blocks, while the second is a Mini-ResNet-inspired CNN with enhanced feature extraction capability. Both models aim to balance classification accuracy and computational efficiency, enabling deployment on resource-limited devices. Experimental results demonstrate that the Mini-ResNet model achieves ۹۲% accuracy, while the simple CNN achieves ۸۸%. Our findings suggest that these lightweight architectures are promising candidates for real-time skin cancer screening applications.

Keywords:

Convolutional Neural Network , Lightweight CNN , Mini-ResNet , PH۲ Dataset , Skin Cancer