Skin cancer is an important and potentially fatal medical disorder brought on by the unnatural proliferation of skin cells. Early identification and exact detection are important for efficient treatment and enhanced patient outcomes. However, subjective skin cancer detection can be laborious and subject to person-to-person variability, calling for the creation of automated and objective diagnostic methods. The deep learning method was applied, and the accuracy of the models was compared for efficient skin cancer detection. Different models such as convolutional neural networks and Alexnet architecture were implemented on the HAM10000 database. Quantitative calculation metrics, including F1 score and accuracy, are used to examine classification accuracy. The results display the upper hand of the suggested method in efficiently detecting cancerous areas under different challenging situations, such as the presence of hair, blurriness in the image, and the presence of different objects. The presented method performs better compared to existing methods in terms of classification accuracy, that is, 92.05% in diverse skin image sources using the pre-trained ResNet50 model.