Latent-based Diffusion Model for Long-tailed Recognition

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Han, Pengxiao, Ye, Changkun, Zhou, Jieming, Zhang, Jing, Hong, Jie, Li, Xuesong
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913324619268096
author Han, Pengxiao
Ye, Changkun
Zhou, Jieming
Zhang, Jing
Hong, Jie
Li, Xuesong
author_facet Han, Pengxiao
Ye, Changkun
Zhou, Jieming
Zhang, Jing
Hong, Jie
Li, Xuesong
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2404_04517
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Latent-based Diffusion Model for Long-tailed Recognition
Han, Pengxiao
Ye, Changkun
Zhou, Jieming
Zhang, Jing
Hong, Jie
Li, Xuesong
Computer Vision and Pattern Recognition
Artificial Intelligence
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.
title Latent-based Diffusion Model for Long-tailed Recognition
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2404.04517