Skin Cancer Image Generation Using WGAN-GP Based on HAM10000 Dataset

DOI

10.2991/978-94-6239-648-7_66How to use a DOI?

Keywords

Skin Lesion Image; WGAN-GP; HAM10000 Dataset; Medical Image Synthesis; Data Augmentation

Abstract

Skin cancer is a highly prevalent disease, and its early diagnosis relies on the recognition of skin lesion images. However, the acquisition cost of labeled medical image samples is extremely high, which makes it easy for model training to suffer from data imbalance. To address this issue, this paper utilizes the HAM10000 dataset to construct and train a lightweight skin lesion image generation model on the basis of Wasserstein Generative Adversarial Network utilizing Gradient Penalty (WGAN-GP). The study first analyzes and preprocesses the HAM10000 dataset, including image standardization and label encoding. Subsequently, a Convolutional Neural Network (CNN) model and a multi-layer perceptron generator are constructed. By increasing the latent variable dimension to 512, setting different specific rates for the generator and discriminator, and introducing gradient penalty and multi-step discriminator update strategies, the training stability and balance are improved. Through adversarial training, high-quality skin lesion images are generated, and the quality of the generated skin lesion images is verified using the Fréchet Inception Distance (FID) score. Experimental results show that the generation model can synthesize reasonably distributed skin lesion images after 5000 steps of training, providing effective support for sample expansion of subsequent classification models. The research in this paper can provide a new direction for the rapid generation of images in the medical imaging field and lay a foundation for the improvement of clinical diagnostic assistance systems.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Hongye Hao
PY  - 2026
DA  - 2026/04/24
TI  - Skin Cancer Image Generation Using WGAN-GP Based on HAM10000 Dataset
BT  - Proceedings of the International Workshop on Advances in Deep Learning for Image Analysis and Computer Vision (IWADIC 2025)
PB  - Atlantis Press
SP  - 605
EP  - 619
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6239-648-7_66
DO  - 10.2991/978-94-6239-648-7_66
ID  - Hao2026
ER  -