Skin diseases are a major public health concern, demanding accurate and timely diagnosis. This study evaluates the performance of five pre-trained deep learning models - VGG16, ResNet50, EfficientNet B0, DenseNet121, and Inception v3 - for skin disease classification using a publicly available dataset. The models were fine-tuned and compared based on accuracy, precision and recall. Our findings demonstrate the potential of these models for supporting dermatologists in clinical practice. However, challenges such as data imbalance and model interpretability remain. Future research should focus on addressing these limitations and exploring ensemble methods for further performance enhancement.