A Re-Parameterized Vision Transformer (ReVT) for Domain-Generalized Semantic Segmentation

被引:0
|
作者
Termoehlen, Jan-Aike [1 ]
Bartels, Timo [1 ]
Fingscheidt, Tim [1 ]
机构
[1] Tech Univ Carolo Wilhelmina Braunschweig, Braunschweig, Germany
关键词
D O I
10.1109/ICCVW60793.2023.00472
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The task of semantic segmentation requires a model to assign semantic labels to each pixel of an image. However, the performance of such models degrades when deployed in an unseen domain with different data distributions compared to the training domain. We present a new augmentation-driven approach to domain generalization for semantic segmentation using a re-parameterized vision transformer (ReVT) with weight averaging of multiple models after training. We evaluate our approach on several benchmark datasets and achieve state-of-the-art mIoU performance of 47.3% (prior art: 46.3%) for small models and of 50.1% (prior art: 47.8%) for midsized models on commonly used benchmark datasets. At the same time, our method requires fewer parameters and reaches a higher frame rate than the best prior art. It is also easy to implement and, unlike network ensembles, does not add any computational complexity during inference.
引用
收藏
页码:4378 / 4387
页数:10
相关论文
共 50 条
  • [1] Reformer: Re-parameterized kernel lightweight transformer for grape disease segmentation
    Zhang, Xinxin
    Feng, Zibo
    Mu, Weisong
    EXPERT SYSTEMS WITH APPLICATIONS, 2025, 265
  • [2] RepMLPNet: Hierarchical Vision MLP with Re-parameterized Locality
    Ding, Xiaohan
    Chen, Honghao
    Zhang, Xiangyu
    Han, Jungong
    Ding, Guiguang
    2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022), 2022, : 568 - 577
  • [3] A Patch Diversity Transformer for Domain Generalized Semantic Segmentation
    He, Pei
    Jiao, Licheng
    Shang, Ronghua
    Liu, Xu
    Liu, Fang
    Yang, Shuyuan
    Zhang, Xiangrong
    Wang, Shuang
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2024, 35 (10) : 14138 - 14150
  • [4] A unified deep semantic expansion framework for domain-generalized person re-identification
    Ang, Eugene P. W.
    Lin, Shan
    Kot, Alex C.
    NEUROCOMPUTING, 2024, 600
  • [5] HGFormer: Hierarchical Grouping Transformer for Domain Generalized Semantic Segmentation
    Ding, Jian
    Xue, Nan
    Xia, Gui-Song
    Schiele, Bernt
    Dai, Dengxin
    2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2023, : 15413 - 15423
  • [6] Image Denoise Model Based on Structural Re-Parameterized Uniformer Transformer and UNet
    Lu, Zhengwei
    Zhang, Duzhen
    Wang, Tao
    Jiang, Hengchang
    Pu, Yingkai
    INTERNATIONAL JOURNAL OF COMPUTATIONAL INTELLIGENCE AND APPLICATIONS, 2024, 23 (02)
  • [7] Hybrid attention transformer with re-parameterized large kernel convolution for image super-resolution
    Ma, Zhicheng
    Liu, Zhaoxiang
    Wang, Kai
    Lian, Shiguo
    IMAGE AND VISION COMPUTING, 2024, 149
  • [8] Adversarial Semantic Hallucination for Domain Generalized Semantic Segmentation
    Tjio, Gabriel
    Liu, Ping
    Zhou, Joey Tianyi
    Goh, Rick Siow Mong
    2022 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2022), 2022, : 3849 - 3858
  • [9] Laformer: Vision Transformer for Panoramic Image Semantic Segmentation
    Yuan, Zheng
    Wang, Junhua
    Lv, Yuxin
    Wang, Ding
    Fang, Yi
    IEEE SIGNAL PROCESSING LETTERS, 2023, 30 : 1792 - 1796
  • [10] DAT: DOMAIN ADAPTIVE TRANSFORMER FOR DOMAIN ADAPTIVE SEMANTIC SEGMENTATION
    Park, Jinyoung
    Son, Minseok
    Lee, Sumin
    Kim, Changick
    2022 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, ICIP, 2022, : 4183 - 4187