Scholarlink

One of the most prevalent cancers in the world is skin cancer. Timely treatment and favorable patient outcomes depend heavily on early detection and precise diagnosis. However, detecting skin cancer can be challenging and requires a dermatologist with extensive training. By categorizing skin cancer, CNNs, one kind of deep learning model, may be able to detect and diagnose it sooner. This work made use of the HAM10000 dataset, which provides numerous high-quality dermatoscopic images of various skin lesions. The proposed InceptionV3 + SVM model in this study is a deep learning model with very good skin cancer classification performance. It has many convolutional, pooling, and thick layers. The training data is oversampled to address the class imbalance. The model was compared to several ML classification techniques, including random forest, DBN, CNN, KNN, and SVM, as well as feature extraction models, including ResNet50, VGG-16, EfficientNet, and DensNet. The model learns the complex patterns and features seen in the images by training for 60 epochs with a batch size of 128. This improves classification performance. With a precision of 97.1%, the accuracy of our suggested model is 97.3%. Utilizing the model checkpoint call-back guarantees that the optimal model is preserved for upcoming applications and replication.

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

Implementation InceptionV3 + SVM Model to Achieve High Accuracy for Skin Cancer Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Anushree Raj,
  • M. Yogitha,
  • Shamitha S. Hegde

摘要

One of the most prevalent cancers in the world is skin cancer. Timely treatment and favorable patient outcomes depend heavily on early detection and precise diagnosis. However, detecting skin cancer can be challenging and requires a dermatologist with extensive training. By categorizing skin cancer, CNNs, one kind of deep learning model, may be able to detect and diagnose it sooner. This work made use of the HAM10000 dataset, which provides numerous high-quality dermatoscopic images of various skin lesions. The proposed InceptionV3 + SVM model in this study is a deep learning model with very good skin cancer classification performance. It has many convolutional, pooling, and thick layers. The training data is oversampled to address the class imbalance. The model was compared to several ML classification techniques, including random forest, DBN, CNN, KNN, and SVM, as well as feature extraction models, including ResNet50, VGG-16, EfficientNet, and DensNet. The model learns the complex patterns and features seen in the images by training for 60 epochs with a batch size of 128. This improves classification performance. With a precision of 97.1%, the accuracy of our suggested model is 97.3%. Utilizing the model checkpoint call-back guarantees that the optimal model is preserved for upcoming applications and replication.