Generative Active Learning for Long-tailed Instance Segmentation

Muzhi Zhu, Chengxiang Fan, Hao Chen, Yang Liu, Weian Mao, Xiaogang Xu, Chunhua Shen
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:62349-62368, 2024.

Abstract

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 estimates the contribution of the current batch-generated data based on gradient cache. BSGAL is meticulously designed to cater for unlimited generated data and complex downstream segmentation tasks. BSGAL outperforms the baseline approach and effectually improves the performance of long-tailed segmentation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-zhu24b, title = {Generative Active Learning for Long-tailed Instance Segmentation}, author = {Zhu, Muzhi and Fan, Chengxiang and Chen, Hao and Liu, Yang and Mao, Weian and Xu, Xiaogang and Shen, Chunhua}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {62349--62368}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/zhu24b/zhu24b.pdf}, url = {https://proceedings.mlr.press/v235/zhu24b.html}, abstract = {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 estimates the contribution of the current batch-generated data based on gradient cache. BSGAL is meticulously designed to cater for unlimited generated data and complex downstream segmentation tasks. BSGAL outperforms the baseline approach and effectually improves the performance of long-tailed segmentation.} }
Endnote
%0 Conference Paper %T Generative Active Learning for Long-tailed Instance Segmentation %A Muzhi Zhu %A Chengxiang Fan %A Hao Chen %A Yang Liu %A Weian Mao %A Xiaogang Xu %A Chunhua Shen %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-zhu24b %I PMLR %P 62349--62368 %U https://proceedings.mlr.press/v235/zhu24b.html %V 235 %X 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 estimates the contribution of the current batch-generated data based on gradient cache. BSGAL is meticulously designed to cater for unlimited generated data and complex downstream segmentation tasks. BSGAL outperforms the baseline approach and effectually improves the performance of long-tailed segmentation.
APA
Zhu, M., Fan, C., Chen, H., Liu, Y., Mao, W., Xu, X. & Shen, C.. (2024). Generative Active Learning for Long-tailed Instance Segmentation. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:62349-62368 Available from https://proceedings.mlr.press/v235/zhu24b.html.

Related Material