Acta Optica Sinica, Volume. 45, Issue 15, 1528001(2025)

Remote Sensing Image Classification Based on Grouped Spatial Coordinate Attention and Mamba

Hui Chen and Zixu Li*
Author Affiliations
  • School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001, Anhui , China
  • show less
    Figures & Tables(12)
    Architecture of Mamba
    Architecture of GCDM-Mamba
    GSCA module
    Architecture of dual-stream multi-directional Mamba encoder
    Some example images of the dataset. (a) Sample images of UCM dataset; (b) sample images of AID dataset; (c) sample images of RESISC45 dataset
    Comparison of Params and FLOPs among different methods
    Influence of the class token position
    • Table 1. Performance comparison of image classification methods on the UCM dataset

      View table

      Table 1. Performance comparison of image classification methods on the UCM dataset

      MethodP /%R /%F1 /%
      ResNet-5091.9091.8491.79
      EfficientNet-B389.9188.9889.12
      ViT-B91.0990.7990.77
      Swin-B91.8591.7491.62
      ViM92.9292.5492.43
      VMamba93.6693.3393.08
      RSMamba-H95.4795.2395.25
      Ours97.2897.1597.13
    • Table 2. Performance comparison of image classification methods on the AID dataset

      View table

      Table 2. Performance comparison of image classification methods on the AID dataset

      MethodP /%R /%F1 /%
      ResNet-5089.8388.8988.62
      EfficientNet-B387.8087.1087.03
      ViT-B88.7188.2688.30
      Swin-B89.8689.3689.39
      ViM90.8190.4290.48
      VMamba92.5891.7591.76
      RSMamba-H92.9792.5192.63
      Ours94.6094.3594.41
    • Table 3. Performance comparison of image classification methods on the NWPU-RESISC45 dataset

      View table

      Table 3. Performance comparison of image classification methods on the NWPU-RESISC45 dataset

      MethodP /%R /%F1 /%
      ResNet-5092.5892.2192.21
      EfficientNet-B389.2188.2587.52
      ViT-B89.3688.2587.88
      Swin-B93.6391.5893.56
      ViM93.0392.8992.86
      VMamba94.3294.2594.22
      RSMamba-H95.2295.1995.18
      Ours96.4196.3596.33
    • Table 4. Validation of the effectiveness of each module

      View table

      Table 4. Validation of the effectiveness of each module

      MethodUCMAIDNWPU-RESISC45
      PRF1PRF1PRF1
      BL92.9292.5492.4390.8190.4290.4893.0392.8992.86
      BL+GSCA95.1894.7694.6693.0392.6592.6195.3595.3095.29
      BL+DMME94.6794.4494.3792.6692.3392.3394.5594.4794.44
      Ours97.2897.1597.1394.6094.3594.4196.4196.3596.33
    • Table 5. Effectiveness verification of the combination of spatial information extraction and attention mechanism

      View table

      Table 5. Effectiveness verification of the combination of spatial information extraction and attention mechanism

      MethodF1 /%
      UCMAIDNWPU-RESISC45
      None94.3792.3394.44
      CA95.2393.1395.32
      GSCA97.1394.4196.33
    Tools

    Get Citation

    Copy Citation Text

    Hui Chen, Zixu Li. Remote Sensing Image Classification Based on Grouped Spatial Coordinate Attention and Mamba[J]. Acta Optica Sinica, 2025, 45(15): 1528001

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Remote Sensing and Sensors

    Received: Apr. 17, 2025

    Accepted: May. 12, 2025

    Published Online: Aug. 8, 2025

    The Author Email: Zixu Li (1074301430@qq.com)

    DOI:10.3788/AOS250956

    CSTR:32393.14.AOS250956

    Topics