SkinNet++: A Lightweight Ensemble Convolutional Layers for Real-Time Skin Cancer Detection

In metro cities, many people face skin issues because of air pollution, changes in water, and synthetic food, mainly affecting the outside layer, which causes itching and swelling. Among all the skin diseases, skin cancer is one of the most common and life-threatening types of cancer, and early detection can have a significant impact on patient outcomes. In this paper, we introduce SkinNet++, a Novel Lightweight Multi-Layered Learning Model (NLM-LLM) based on an ensemble of convolutional layers designed to be efficient and accurate for real-time skin cancer detection. The existing models are more overloaded architectures that obtain memory issues. The pretrained model Adaptive InceptionV3 was trained on various benchmark datasets. InceptionV3 can capture complex skin lesions based on the affected patterns using RGB images. SkinNet++ adopts multiple shallow and optimized convolutional blocks for extracting diverse feature representations while introducing low computational costs with advanced classification rates. The model is trained and tested on benchmark datasets ISIC and HAM10000, and results compare favorably in accuracy of 98.56 %, sensitivity of 99.34, and specificity of 99.41. With its lightweight architecture and high inference speed, it is also applied to mobile and edge devices where it is possible to provide immediate support for medical diagnosis, even in resource-scarce conditions.