Lightweight CNN-Attention Achieves 87% Accuracy in Skin Cancer Detection
October 28, 2026
A lightweight attention-based CNN achieves 87% accuracy and 0.9475 ROC AUC for binary skin lesion classification using only 1.95 million parameters. The model matches MobileNet performance while requiring approximately half the parameters and storage, making it suitable for edge deployment.
HOW THIS AFFECTS YOU
●
builderYou can deploy this low-parameter model for mobile or embedded medical imaging applications.
●
healthThis offers a highly efficient diagnostic tool for early skin cancer detection in resource-constrained environments.