This repository contains the testing code and pre-trained models for the paper Calibrated Hyperspectral Image Reconstruction via Graph-based Self-Tuning Network.
- Python 3.7.10
- Pytorch 1.9.1
- Numpy 1.21.2
- Scipy 1.7.1
Ten simulation testing HSI (256x256x28) are provided. Testing trials can be determined by specify trial_num
To test a pre-trained model under miscalibration many-to-many, specify mode as many_to_many, last_train as 661.
To test a pre-trained model under miscalibration one-to-many, specify mode as one_to_many, last_train as 662.
To test a pre-trained model under traditional setting one-to-one, specify mode as one_to_one, last_train as 662.
Run
python test.py
| directory | description |
|---|---|
Data |
Ten simulation testing HSIs and two real masks for testing (256x256 and 660x660) |
test |
testing script |
utils |
utility functions |
tools |
model components |
ssim_torch |
function for computing SSIM |
many_to_many |
model structure and model checkpoint for miscalibration (many-to-many) |
one_to_many |
model structure and model checkpoint for miscalibration (one-to-many) |
one_to_one |
model structure and model checkpoint for traditional setting (one-to-one) |