MobDensNet : Advanced Hybrid deep learning model for Multiclass Skin Cancer Classification

Early identification and treatment of skin cancer greatly improves survival rates. Recent improvements in deep learning algorithms have showed promise for detecting skin cancer. This study provides many novel deep learning-based frameworks for multi classification of skin cancer. The Advance RDUNet model combines convolutional neural networks (CNN) with three cutting-edge pre-trained classifiers in the medical area, which were assessed using the HAM10000 dataset and achieved higher multiclassification accuracy than conventional techniques. In addition, MobileNet-V3 and DenseNet deep neural network (MobDenseNet) is created for skin lesion classification, which uses Ensemble learning that utilizes pooling, and normalizing to improve picture quality. Deep features were retrieved with a residual neural network leveraging transfer learning and global average pooling and trained with Extreme Learning with ReLu activation function. On the HAM10000 datasets, MobDenseNet model. Furthermore, this hybrid model obtained 97.56% accuracy; hence, the suggested approach can aid expert clinicians in auto matically identifying cancer based on skin lesion images.