The 2-nd place solution, which won the runner-up award in PIRM spectral super-resolution challenge. The repo now contains the testing codes which enable user to validate/test the pretrained model on testing/validation set
The main dependencies are listed as follows (others are left but should also be installed if missed)
conda install pytorch=0.4.0
pip install torchnet torchvision
The pretrained models could be downloaded from this path, which contains the EMSR and EMSR-CA models we submitted in final testing stage of PIRM2018-Spectral-SR challenge.
The procedures of testing our pretrained models are listed as follows:
-
use envi2mat.m to transform your envi data into mat data
-
modify the pathes in hsi_test.py
-
finally, run the code below
python hsi_test.py -a emsrx3 -r -rp /path/to/emsrx3/model_best.pth --sf 3 --self-ensemble --test --no-log -nro
python hsi_test.py -a emsrcax3 -r -rp /path/to/emsrcax3/model_best.pth --sf 3 --self-ensemble --test --no-log -nro
If you have any questions, please do not hesitate to contact me (kaixuan_wei@outlook.com).