Melanoma (MEL), squamous cell carcinoma (SCC), and basal cell carcinoma (BCC) are among the skin cancers that have become increasingly prevalent and are a danger to world health. The effective treatment and diagnosis of these conditions are reliant upon the early and precise identification of the conditions. This study examines the effects of fine-tuning three pre-trained deep learning models—Inception V3, VGG-19, and ResNeXt 50—utilizing the ISIC skin cancer image dataset. To optimize the model's performance, a comprehensive image augmentation strategy is implemented. Fine-tuning involves freezing the initial layers of pre-trained models, replacing the final classification layers with fully connected network (FCN), and adding a SoftMax activation function. The models were assessed utilizing performance parameters including accuracy, precision, recall, and F1 score. Inception V3 attained an accuracy improvement from 92% to 99.8%, while ResNeXt 50 progressed from 91% to 99.82%. VGG-19, %, demonstrated a significant increase of 10.3%, leading to an accuracy of 96.3%.These findings demonstrate that fine-tuning deep learning models for medical diagnostics can increase skin cancer diagnosis accuracy and patient care.