Hyperspectral Target Detection Based on Interpretable Representation Network (HTD-IRN), IEEE TGRS, 2023.

Files

Main.py, Model.py, Train_Test.py,ts_generation.py,utils.py.

Requirement

python 3.6.13, torch 1.10.1, NVIDIA Geforce RTX 2080Ti.

Network

HTD-IRN is a promising detector for hyperspectral imagery based on a deep subspace representation network with Uformer.

Run

run Main.py.

Note

1. Three states of train, test, or parameter_selection can be chosen in Main.py.

2. We provide a well-trained model of San Diego I, and you can test it directly.

3. For a new dataset, the optimal values of m and eta1 should be chosen first. So you can change the state to parameter_selection in Main.py.

Cite

@ARTICLE{shen2023hyperspectral,
  author={Shen, Dunbin and Ma, Xiaorui and Kong, Wenfeng and Liu, Jianjun and Wang, Jie and Wang, Hongyu},
  journal={IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING}, 
  title={Hyperspectral Target Detection Based on Interpretable Representation Network}, 
  year={2023},
  volume={},
  number={},
  pages={1-17},
  doi={10.1109/TGRS.2023.3302950}}