Scholarlink

Detecting skin cancer through skin pigment lesions is a topic of significant interest. In the paper, we propose deep-learning techniques for classifying skin cancer. We use the HAM10000 dataset with 10,015 images belonging to 7 different diseases. Additionally, to balance the data, we collect more instances and employ various data augmentation techniques. We propose models such as ResNet, VGG16, and AlexNet and enhance these models to achieve the best results. The resulting accuracies are 92.9%, 98.4%, and 88.3%, respectively. Based on comparison and evaluation results, we selected the improved VGG16 model for the skin cancer classification problem. Hopefully, this research can aid in the early detection and timely treatment of pathogens.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving Architectures of VGG16, AlexNet, and ResNet50 Models for Skin Cancer Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Dau Sy Hieu,
  • Dang Thi Phuc,
  • Le Ngoc Ton,
  • Tran Chi Hung,
  • Nguyen Cao Cuong

摘要

Detecting skin cancer through skin pigment lesions is a topic of significant interest. In the paper, we propose deep-learning techniques for classifying skin cancer. We use the HAM10000 dataset with 10,015 images belonging to 7 different diseases. Additionally, to balance the data, we collect more instances and employ various data augmentation techniques. We propose models such as ResNet, VGG16, and AlexNet and enhance these models to achieve the best results. The resulting accuracies are 92.9%, 98.4%, and 88.3%, respectively. Based on comparison and evaluation results, we selected the improved VGG16 model for the skin cancer classification problem. Hopefully, this research can aid in the early detection and timely treatment of pathogens.