search

radimspetlik / SI-DDPM-FMO

[WACV 2024] SI-DDPM-FMO: Single-Image Deblurring, Trajectory and Shape Recovery of Fast Moving Objects with Denoising Diffusion Probabilistic Models
https://github.com/radimspetlik/SI-DDPM-FMO
MIT License
3 stars 1 forks source link

How to deblur a single motion blurred image? #2

Open jiachen0212 opened 3 weeks ago

jiachen0212 commented 3 weeks ago

Thanks for your work. I want to use code to deblur a single motion blurred image. I have studied the fmo_deblurring_benchmark.py script, and it seems that it needs the ground truth information of the gt_bbox box. How can I modify it to skip this supervision information and deblur a single motion blurred image?

rasemailcz commented 3 weeks ago

Dear @jiachen0212 ,

The bounding box is there since it is required by the benchmarks. You actually do not need to provide a bounding box.

The deblurring and trajectory and shape estimation happens from the line 166 in fmo_deblurring_benchmark.py https://github.com/radimspetlik/SI-DDPM-FMO/blob/217a300bf7999d8518468e3bc7c272f96aa1a3ab/fmo_deblurring_benchmark.py#L166 and the image sequence is created on the line 247 https://github.com/radimspetlik/SI-DDPM-FMO/blob/217a300bf7999d8518468e3bc7c272f96aa1a3ab/fmo_deblurring_benchmark.py#L247.

Hope this helps...