“Skin cancer” is a prevalent and potentially life-threatening condition that arises from the uncontrolled growth of abnormal skin cells. It is crucial to identify and classify “skin cancer” accurately for effective “diagnosis” and “treatment.” By leveraging dense nets and advanced algorithms, deep learning models can be used to accurately analyse dermatoscopic images. The additional convolution layer is used here to modify the “DenseNet” model to detect “skin cancer” more accurately. A thorough comparison is made between the modified “DenseNet,” “ResNet,” “Xception,” models, and “MobileNet.” The superiority of our modified structure is proven. The modified “DenseNet-201” model achieves more than 95% accuracy compared with other techniques.