Pytorch code for "BSRT: Improving Burst Super-Resolution with Swin Transformer and Flow-Guided Deformable Alignment", CVPRW, 1st place in NTIRE 2022 BurstSR Challenge (real-world track).
The data-processing module should be refactored into its own, standalone module outside the BSRT module. It offers a number of utility functions, like creating RAW bursts from a single image, which are very useful.
Hi @ConnorBaker thanks for sharing your code.
Regarding the hyperparameters searching, do you intend to search a lightweight architecture? how about the resulted model in size/param/flops? thanks