AR-TTA: A Simple Method for Real-World Continual Test-Time Adaptation

被引:0
|
作者
Sojka, Damian [1 ,2 ]
Cygert, Sebastian [2 ,3 ]
Twardowski, Bartlomiej [2 ,4 ,5 ]
Trzcinski, Tomasz [2 ,6 ,7 ]
机构
[1] Poznan Univ Tech, Poznan, Poland
[2] IDEAS NCBR, Warsaw, Poland
[3] Gdansk Univ Technol, Gdansk, Poland
[4] Autonomous Univ Barcelona, Barcelona, Spain
[5] Comp Vis Ctr, Barcelona, Spain
[6] Warsaw Univ Technol, Warsaw, Poland
[7] Tooploox, Wroclaw, Poland
关键词
D O I
10.1109/ICCVW60793.2023.00374
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Test-time adaptation is a promising research direction that allows the source model to adapt itself to changes in data distribution without any supervision. Yet, current methods are usually evaluated on benchmarks that are only a simplification of real-world scenarios. Hence, we propose to validate test-time adaptation methods using the recently introduced datasets for autonomous driving, namely CLAD-C and SHIFT. We observe that current test-time adaptation methods struggle to effectively handle varying degrees of domain shift, often resulting in degraded performance that falls below that of the source model. We noticed that the root of the problem lies in the inability to preserve the knowledge of the source model and adapt to dynamically changing, temporally correlated data streams. Therefore, we enhance well-established self-training framework by incorporating a small memory buffer to increase model stability and at the same time perform dynamic adaptation based on the intensity of domain shift. The proposed method, named AR-TTA, outperforms existing approaches on both synthetic and more real-world benchmarks and shows robustness across a variety of TTA scenarios.
引用
收藏
页码:3483 / 3487
页数:5
相关论文
共 39 条
  • [1] Continual Test-Time Domain Adaptation
    Wang, Qin
    Fink, Olga
    Van Gool, Luc
    Dai, Dengxin
    [J]. 2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2022, : 7191 - 7201
  • [2] RDumb: A simple approach that questions our progress in continual test-time adaptation
    Press, Ori
    Schneider, Steffen
    Kummerer, Matthias
    Bethge, Matthias
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 36 (NEURIPS 2023), 2023,
  • [3] Multiple Teacher Model for Continual Test-Time Domain Adaptation
    Wang, Ran
    Zuo, Hua
    Fang, Zhen
    Lu, Jie
    [J]. ADVANCES IN ARTIFICIAL INTELLIGENCE, AI 2023, PT I, 2024, 14471 : 304 - 314
  • [4] Robust Mean Teacher for Continual and Gradual Test-Time Adaptation
    Doebler, Mario
    Marsden, Robert A.
    Yang, Bin
    [J]. 2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR, 2023, : 7704 - 7714
  • [5] Exploring Safety Supervision for Continual Test-time Domain Adaptation
    Yang, Xu
    Gu, Yanan
    Wei, Kun
    Deng, Cheng
    [J]. PROCEEDINGS OF THE THIRTY-SECOND INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, IJCAI 2023, 2023, : 1649 - 1657
  • [6] Noise-Robust Continual Test-Time Domain Adaptation
    Yu, Zhiqi
    Li, Jingjing
    Du, Zhekai
    Li, Fengling
    Zhu, Lei
    Yang, Yang
    [J]. PROCEEDINGS OF THE 31ST ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA, MM 2023, 2023, : 2654 - 2662
  • [7] NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation
    Gong, Taesik
    Jeong, Jongheon
    Kim, Taewon
    Kim, Yewon
    Shin, Jinwoo
    Lee, Sung-Ju
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022), 2022,
  • [8] Towards Real-World Test-Time Adaptation: Tri-net Self-Training with Balanced Normalization
    Su, Yongyi
    Xu, Xun
    Jia, Kui
    [J]. THIRTY-EIGHTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 38 NO 13, 2024, : 15126 - 15135
  • [9] Ev-TTA: Test-Time Adaptation for Event-Based Object Recognition
    Kim, Junho
    Hwang, Inwoo
    Kim, Young Min
    [J]. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2022, 2022-June : 17724 - 17733
  • [10] Diffusion-TTA: Test-time Adaptation of Discriminative Models via Generative Feedback
    Prabhudesai, Mihir
    Ke, Tsung-Wei
    Li, Alexander C.
    Pathak, Deepak
    Fragkiadaki, Katerina
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 36 (NEURIPS 2023), 2023,