PyTorch code for paper "Dual-Diffusion: Dual Conditional Denoising Diffusion Probabilistic Models for Blind Super-Resolution Reconstruction in RSIs", which can be seen at https://doi.org/10.48550/arXiv.2305.12170. The code is based on https://github.com/megvii-research/DCLS-SR/tree/master/codes
- RRDB_LR encoder (The pretrained RRDB_LR encoder has been given in the folder "Pretrained rrdb_LR encoder")
- Kernel Predictor
- HR reconstructor
The link to the dataset: https://pan.baidu.com/s/1eD_mbFoNdPWfY8TCkjjfeA?pwd=j3vq
To transform datasets to binary files for efficient IO, run:
python codes/scripts/create_lmdb.py
To generate LRblur/LR_up/Bicubic datasets paths, run:
python codes/scripts/generate_mod_blur_LR_bic.py (You need to modify the file paths by yourself.)
-
How good is the proposed HR reconstructor compared to non-blind methods when all those methods are given the same (predicted or true) kernel？
-
How good is the kernel predictor compared to the other blind methods' predictors
