Published Paper Details:

Skin cancer, one of the most prevalent cancers worldwide, requires timely and accurate detection for better pa- tient outcomes. This project investigates the use of deep learning methods for automatically classifying skin cancer from dermoscopic pictures. Specifically, convolutional neural networks (CNNs) and hybrid models incorporating transfer learning are employed to enhance classification accuracy. To tackle issues including unequal class distribution, noise in images, and restricted data accessibility, preprocessing methods like data augmentation, denoising, and feature extraction are implemented. The project explores novel model architectures, including compact transformers and dual autoencoder models, aiming to improve the model's performance. Additionally, established benchmark datasets such as HAM10000 and ISIC-2019 are used for training and evaluation. The effective- ness of the developed model is assessed using key metrics, includ- ing accuracy, sensitivity, and specificity. The results highlight the potential of deep learning in accurately classifying skin lesions as malignant or benign, offering valuable support for clinicians in early skin cancer diagnosis, especially in resource-limited settings.By highlighting current trends, obstacles, and potential avenues for future research in the field, this work advances automated diagnostic systems for skin cancer