AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach

被引:9
|
作者
Song, Xiao [1 ]
Yang, Guorun [2 ]
Zhu, Xinge [3 ]
Zhou, Hui [1 ]
Ma, Yuexin [4 ]
Wang, Zhe [1 ]
Shi, Jianping [1 ]
机构
[1] SenseTime Grp Ltd, Hong Kong, Peoples R China
[2] Chinese Acad Sci, Shenzhen Inst Adv Technol, Beijing, Peoples R China
[3] Chinese Univ Hong Kong, Hong Kong, Peoples R China
[4] ShanghaiTech Univ, Shanghai, Peoples R China
关键词
Domain adaptation; Stereo matching; Non-adversarial progressive color transfer; Cost normalization; Self-supervised occlusion-aware reconstruction;
D O I
10.1007/s11263-021-01549-6
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Recently, records on stereo matching benchmarks are constantly broken by end-to-end disparity networks. However, the domain adaptation ability of these deep models is quite limited. Addressing such problem, we present a novel domain-adaptive approach called AdaStereo that aims to align multi-level representations for deep stereo matching networks. Compared to previous methods, our AdaStereo realizes a more standard, complete and effective domain adaptation pipeline. Firstly, we propose a non-adversarial progressive color transfer algorithm for input image-level alignment. Secondly, we design an efficient parameter-free cost normalization layer for internal feature-level alignment. Lastly, a highly related auxiliary task, self-supervised occlusion-aware reconstruction is presented to narrow the gaps in output space. We perform intensive ablation studies and break-down comparisons to validate the effectiveness of each proposed module. With no extra inference overhead and only a slight increase in training complexity, our AdaStereo models achieve state-of-the-art cross-domain performance on multiple benchmarks, including KITTI, Middlebury, ETH3D and DrivingStereo, even outperforming some state-of-the-art disparity networks finetuned with target-domain ground-truths. Moreover, based on two additional evaluation metrics, the superiority of our domain-adaptive stereo matching pipeline is further uncovered from more perspectives. Finally, we demonstrate that our method is robust to various domain adaptation settings, and can be easily integrated into quick adaptation application scenarios and real-world deployments.
引用
收藏
页码:226 / 245
页数:20
相关论文
共 50 条
  • [1] AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach
    Xiao Song
    Guorun Yang
    Xinge Zhu
    Hui Zhou
    Yuexin Ma
    Zhe Wang
    Jianping Shi
    [J]. International Journal of Computer Vision, 2022, 130 : 226 - 245
  • [2] Correction to: AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach
    Xiao Song
    Guorun Yang
    Xinge Zhu
    Hui Zhou
    Yuexin Ma
    Zhe Wang
    Jianping Shi
    [J]. International Journal of Computer Vision, 2022, 130 : 884 - 884
  • [3] AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach (vol 130, pg 226, 2022)
    Song, Xiao
    Yang, Guorun
    Zhu, Xinge
    Zhou, Hui
    Ma, Yuexin
    Wang, Zhe
    Shi, Jianping
    [J]. INTERNATIONAL JOURNAL OF COMPUTER VISION, 2022, 130 (03) : 884 - 884
  • [4] AdaStereo: A Simple and Efficient Approach for Adaptive Stereo Matching
    Song, Xiao
    Yang, Guorun
    Zhu, Xinge
    Zhou, Hui
    Wang, Zhe
    Shi, Jianping
    [J]. 2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, : 10323 - 10332
  • [5] Domain-adaptive modules for stereo matching network
    Ling, Zhi
    Yang, Kai
    Li, Jinlong
    Zhang, Yu
    Gao, Xiaorong
    Luo, Lin
    Xie, Liming
    [J]. NEUROCOMPUTING, 2021, 461 : 217 - 227
  • [6] Domain-Adaptive 3D Medical Image Synthesis: An Efficient Unsupervised Approach
    Hu, Qingqiao
    Li, Hongwei
    Zhang, Jianguo
    [J]. MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION, MICCAI 2022, PT VI, 2022, 13436 : 495 - 504
  • [7] Efficient Stereo Matching Using Adaptive Manifolds
    Jiang, Lei
    Zhao, Hanli
    [J]. PROCEEDINGS NICOGRAPH INTERNATIONAL 2016, 2016, : 92 - 95
  • [8] Adaptive Neighbor Embedding for Efficient Stereo Matching
    Chong, Ai-Xin
    Yin, Hui
    Wan, Jin
    Liu, Yan-Ting
    Du, Qian-Qian
    [J]. IEEE TRANSACTIONS ON INTELLIGENT VEHICLES, 2024, 9 (01): : 2449 - 2458
  • [9] Generalized Domain-Adaptive Dictionaries
    Shekhar, Sumit
    Patel, Vishal M.
    Nguyen, Hien V.
    Chellappa, Rama
    [J]. 2013 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2013, : 361 - 368
  • [10] COCM: Co-Occurrence-Based Consistency Matching in Domain-Adaptive Segmentation
    Zhu, Siyu
    Tian, Yingjie
    Zhou, Fenfen
    Bai, Kunlong
    Song, Xiaoyu
    [J]. MATHEMATICS, 2022, 10 (23)