Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2504.21302 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909597628891136 |
|---|---|
| author | Shen, Zhelun Li, Zhuo Wu, Chenming Rao, Zhibo Liu, Lina Dai, Yuchao Zhang, Liangjun |
| author_facet | Shen, Zhelun Li, Zhuo Wu, Chenming Rao, Zhibo Liu, Lina Dai, Yuchao Zhang, Liangjun |
| contents | Recently, learning-based stereo matching methods have achieved great improvement in public benchmarks, where soft argmin and smooth L1 loss play a core contribution to their success. However, in unsupervised domain adaptation scenarios, we observe that these two operations often yield multimodal disparity probability distributions in target domains, resulting in degraded generalization. In this paper, we propose a novel approach, Constrain Multi-modal Distribution (CMD), to address this issue. Specifically, we introduce \textit{uncertainty-regularized minimization} and \textit{anisotropic soft argmin} to encourage the network to produce predominantly unimodal disparity distributions in the target domain, thereby improving prediction accuracy. Experimentally, we apply the proposed method to multiple representative stereo-matching networks and conduct domain adaptation from synthetic data to unlabeled real-world scenes. Results consistently demonstrate improved generalization in both top-performing and domain-adaptable stereo-matching models. The code for CMD will be available at: \href{https://github.com/gallenszl/CMD}{https://github.com/gallenszl/CMD}. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_21302 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | CMD: Constraining Multimodal Distribution for Domain Adaptation in Stereo Matching Shen, Zhelun Li, Zhuo Wu, Chenming Rao, Zhibo Liu, Lina Dai, Yuchao Zhang, Liangjun Computer Vision and Pattern Recognition Robotics Recently, learning-based stereo matching methods have achieved great improvement in public benchmarks, where soft argmin and smooth L1 loss play a core contribution to their success. However, in unsupervised domain adaptation scenarios, we observe that these two operations often yield multimodal disparity probability distributions in target domains, resulting in degraded generalization. In this paper, we propose a novel approach, Constrain Multi-modal Distribution (CMD), to address this issue. Specifically, we introduce \textit{uncertainty-regularized minimization} and \textit{anisotropic soft argmin} to encourage the network to produce predominantly unimodal disparity distributions in the target domain, thereby improving prediction accuracy. Experimentally, we apply the proposed method to multiple representative stereo-matching networks and conduct domain adaptation from synthetic data to unlabeled real-world scenes. Results consistently demonstrate improved generalization in both top-performing and domain-adaptable stereo-matching models. The code for CMD will be available at: \href{https://github.com/gallenszl/CMD}{https://github.com/gallenszl/CMD}. |
| title | CMD: Constraining Multimodal Distribution for Domain Adaptation in Stereo Matching |
| topic | Computer Vision and Pattern Recognition Robotics |
| url | https://arxiv.org/abs/2504.21302 |