English

Latent-based Diffusion Model for Long-tailed Recognition

Computer Vision and Pattern Recognition 2024-04-24 v2 Artificial Intelligence

Abstract

Long-tailed imbalance distribution is a common issue in practical computer vision applications. Previous works proposed methods to address this problem, which can be categorized into several classes: re-sampling, re-weighting, transfer learning, and feature augmentation. In recent years, diffusion models have shown an impressive generation ability in many sub-problems of deep computer vision. However, its powerful generation has not been explored in long-tailed problems. We propose a new approach, the Latent-based Diffusion Model for Long-tailed Recognition (LDMLR), as a feature augmentation method to tackle the issue. First, we encode the imbalanced dataset into features using the baseline model. Then, we train a Denoising Diffusion Implicit Model (DDIM) using these encoded features to generate pseudo-features. Finally, we train the classifier using the encoded and pseudo-features from the previous two steps. The model's accuracy shows an improvement on the CIFAR-LT and ImageNet-LT datasets by using the proposed method.

Keywords

Cite

@article{arxiv.2404.04517,
  title  = {Latent-based Diffusion Model for Long-tailed Recognition},
  author = {Pengxiao Han and Changkun Ye and Jieming Zhou and Jing Zhang and Jie Hong and Xuesong Li},
  journal= {arXiv preprint arXiv:2404.04517},
  year   = {2024}
}

Comments

8 pages, 3 figures. Accepted by L3DIVU-CVPR2024

R2 v1 2026-06-28T15:45:46.694Z