Scholarlink

Clothing-related datasets are used to train variational autoencoders, starting with the commonly used Fashion MNIST database. Clustering is applied and analyzed in the latent space and specific points are defined in this latent space for the purpose of generating novel images. These points correspond to the positions in the middle of the lines connecting the centroids of the resulting clusters. These aspects are analyzed with the different datasets focused on fashion images.

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

Clustering the Latent Space in Variational Autoencoders for Image Generation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Diego Martínez,
  • José Santos

摘要

Clothing-related datasets are used to train variational autoencoders, starting with the commonly used Fashion MNIST database. Clustering is applied and analyzed in the latent space and specific points are defined in this latent space for the purpose of generating novel images. These points correspond to the positions in the middle of the lines connecting the centroids of the resulting clusters. These aspects are analyzed with the different datasets focused on fashion images.