/AWAN

Adaptive Weighted Attention Network with Camera Spectral Sensitivity Prior for Spectral Reconstruction from RGB Images in CVPRW 2020 (Winner of NTIRE Challenge, Oral)

Primary LanguagePython

Adaptive Weighted Attention Network with Camera Spectral Sensitivity Prior for Spectral Reconstruction from RGB Images

Jiaojiao Li, Chaoxiong Wu, Rui Song, Yunsong Li and Fei Liu. Adaptive Weighted Attention Network with Camera Spectral Sensitivity Prior for Spectral Reconstruction from RGB Images. In CVPR Workshop 2020. (Winner of NTIRE Challenge on Spectral Reconstruction from RGB Images. Our method achieves the 1st ranking on the "Clean" track and the 3rd place only 1.59106e-4 more than the 1st on the "Real World" track).

AWAN framework

  • Network architecture of the proposed adaptive weighted attention network (AWAN).

AWCA module

  • Diagram of adaptive weighted channel attention (AWCA) module.

PSNL module

  • Diagram of patch-level second-order non-local (PSNL) module.

Citation

  • If the the work or the code is helpful, please cite the paper:

    @InProceedings{Li_2020_CVPR_Workshops,
    author = {Li, Jiaojiao and Wu, Chaoxiong and Song, Rui and Li, Yunsong and Liu, Fei},
    title = {Adaptive Weighted Attention Network With Camera Spectral Sensitivity Prior for Spectral Reconstruction From RGB Images},
    booktitle = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month = {June},
    year = {2020}
    }

CVPRW poster