Skin Cancer Classification using NasNet
Md Masum Billah✉, Amit Deb Natha, Denesh Dasb, Tanvir Mahmudc, Rashedur Rahmand
- aDepartment of Electrical and Electronic Engineering , Leading University , Sylhet , Bangladesh
- bDepartment of Electrical and Electronic Engineering , Southern University Bangladesh , Chattogram , Bangladesh
- cDepartment of EEE , Daffodil International University , Daffodil Smart City (DSC) , Dhaka- 1216 , Dhaka , Bangladesh
- dDepartment of Computer Science and Engineering , Daffodil International University , Dhaka , Bangladesh
- ✉Corresponding author
Papers open inside SciRah from the original publisher or repository; SciRah never hosts them.
Abstract
Skin cancer remains one of the major causes of mortality worldwide, with malignant melanoma being the deadliest type due to its high potential for metastasis. Although relatively uncommon, it accounts for nearly 75% of skin cancer-related deaths. Early detection plays a crucial role in improving outcomes, but it is difficult because melanoma often closely resembles benign skin lesions. In this work, we propose an automated system for early melanoma detection using deep transfer learning. Our method leverages a pre-trained NASNet model, from which features are transferred to a new dataset for classification. We adapted the original network by incorporating global average pooling and customized classification layers. The system was trained and evaluated on skin images from the ISIC 2020 dataset.
TL;DRThis work proposes an automated system for early melanoma detection using deep transfer learning that leverages a pre-trained NASNet model, from which features are transferred to a new dataset for classification. (Semantic Scholar)