Generative Active Learning for Long-tailed Instance Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhu, Muzhi, Fan, Chengxiang, Chen, Hao, Liu, Yang, Mao, Weian, Xu, Xiaogang, Shen, Chunhua
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916273905991680
author Zhu, Muzhi
Fan, Chengxiang
Chen, Hao
Liu, Yang
Mao, Weian
Xu, Xiaogang
Shen, Chunhua
author_facet Zhu, Muzhi
Fan, Chengxiang
Chen, Hao
Liu, Yang
Mao, Weian
Xu, Xiaogang
Shen, Chunhua
contents Recently, large-scale language-image generative models have gained widespread attention and many works have utilized generated data from these models to further enhance the performance of perception tasks. However, not all generated data can positively impact downstream models, and these methods do not thoroughly explore how to better select and utilize generated data. On the other hand, there is still a lack of research oriented towards active learning on generated data. In this paper, we explore how to perform active learning specifically for generated data in the long-tailed instance segmentation task. Subsequently, we propose BSGAL, a new algorithm that online estimates the contribution of the generated data based on gradient cache. BSGAL can handle unlimited generated data and complex downstream segmentation tasks effectively. Experiments show that BSGAL outperforms the baseline approach and effectually improves the performance of long-tailed segmentation. Our code can be found at https://github.com/aim-uofa/DiverGen.
format Preprint
id arxiv_https___arxiv_org_abs_2406_02435
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Generative Active Learning for Long-tailed Instance Segmentation
Zhu, Muzhi
Fan, Chengxiang
Chen, Hao
Liu, Yang
Mao, Weian
Xu, Xiaogang
Shen, Chunhua
Computer Vision and Pattern Recognition
Recently, large-scale language-image generative models have gained widespread attention and many works have utilized generated data from these models to further enhance the performance of perception tasks. However, not all generated data can positively impact downstream models, and these methods do not thoroughly explore how to better select and utilize generated data. On the other hand, there is still a lack of research oriented towards active learning on generated data. In this paper, we explore how to perform active learning specifically for generated data in the long-tailed instance segmentation task. Subsequently, we propose BSGAL, a new algorithm that online estimates the contribution of the generated data based on gradient cache. BSGAL can handle unlimited generated data and complex downstream segmentation tasks effectively. Experiments show that BSGAL outperforms the baseline approach and effectually improves the performance of long-tailed segmentation. Our code can be found at https://github.com/aim-uofa/DiverGen.
title Generative Active Learning for Long-tailed Instance Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2406.02435