Greetings, thank you for sharing your wonderful work. I have re-run the codes and successfully reproduced the results for the Pavia and Botswana datasets. Unfortunately, for the Chikusei dataset, I can only achieve ~35dB instead of 43dB as reported in the paper. I would be very grateful if you can verify the settings in the config files as they may effect the performance. It's worth noting that, for the Pavia and Botswana datasets, I used a single GPU for training. However, for the Chikusei dataset, I used two GPUs as the images are bigger, they cannot fit in a single GPU.
Thank you very much.
Greetings, thank you for sharing your wonderful work. I have re-run the codes and successfully reproduced the results for the Pavia and Botswana datasets. Unfortunately, for the Chikusei dataset, I can only achieve ~35dB instead of 43dB as reported in the paper. I would be very grateful if you can verify the settings in the config files as they may effect the performance. It's worth noting that, for the Pavia and Botswana datasets, I used a single GPU for training. However, for the Chikusei dataset, I used two GPUs as the images are bigger, they cannot fit in a single GPU.
Thank you very much.