Open Jamie-Adams369 opened 1 week ago
For the first question, HSIFormer38 was not utilized in this study, so line#20 can be directly removed.
For the second question, the Houston dataset employed in this research originates from the 2018 IEEE GRSS Data Fusion Contest. We cropped the top region of the image into eight non-overlapping hyperspectral images, each with dimensions of 256 × 256 × 48, for testing. The remaining part of the image was divided into overlapping patches for training, with 10% of the training data randomly selected as a validation set. The low-resolution (LR) hyperspectral images were generated by applying bicubic downsampling to these patches for various scale factors, and are stored under the key "ms" in the test_data
variable. For specific data processing code, please refer to MSDFormer and SSPSR.
Thank you for answering.
Sorry to bother you again.
I wonder if you mind making the processed Houston dataset publicly available.
Sorry to bother you again.
I wonder if you mind making the processed Houston dataset publicly available.
We offer scripts to process the data here: https://huggingface.co/datasets/WHU-Sigma/HyperSIGMA_Datasets/tree/main/HyperSIGMA_super-resolution
Hello, first of all, thank you for sharing your wonderful work.
When reproducing your work on ImageSuperResolution, I encountered two questions and hope to receive answers.
1.It seems that some files are missing from the code you provided.
2.Have you preprocessed the Houston dataset? I am unable to read the data correctly.
Hope for your reply, Thank you!