NeSSR

---

Continuous Spatial-Spectral Reconstruction via Implicit Neural Representation, in International Journal of Computer Vision (IJCV) 2024. Ruikang Xu, Mingde Yao, Chang Chen, Lizhi Wang, and Zhiwei Xiong.

Datesets

Prepare the ICVL dataset for training, it can be downloaded from this link.

Processed test dataset for inference can be downloaded from BaiduYun (Access code: ijcv).

---

Dependencies

• Python 3.8.8, PyTorch 1.8.0, torchvision 0.9.0.
• NumPy 1.24.2, OpenCV 4.7.0, Tensorboardx 2.5.1, Pillow, Imageio, SciPy.

---

Quick Start

Spectral Reconstruction from RGB Images with Arbitrary Spectral Resolution

• Download the pre-trained model from BaiduYun (Access code: ijcv).

• Inference with pre-trained model:

cd ./SpectralRec && python test.py

Spatial Reconstruction from RGB-HSI Pairs with Arbitrary Spatial Resolution

• Download the pre-trained model from BaiduYun (Access code: ijcv).

• Inference with pre-trained model:

cd ./SpatialRec && python test.py

Spatial-Spectral Reconstruction from RGB-MSI Pairs with Arbitrary Spatial-Spectral Resolution

• Download the pre-trained model from BaiduYun (Access code: ijcv).

• Inference with pre-trained model:

cd ./JointRec && python test.py

---

Contact

Any question regarding this work can be addressed to [email protected].

---

Citation

If you find our work helpful, please cite the following paper.

@article{xu2024continuous,
  title={Continuous Spatial-Spectral Reconstruction via Implicit Neural Representation},
  author={Xu, Ruikang and Yao, Mingde and Chen, Chang and Wang, Lizhi and Xiong, Zhiwei},
  journal={International Journal of Computer Vision},
  pages={1--23},
  year={2024},
  publisher={Springer}
}