fast-codi / CoDi

[CVPR24] CoDi: Conditional Diffusion Distillation for Higher-Fidelity and Faster Image Generation
https://fast-codi.github.io/
73 stars 1 forks source link

Any public super-resolution (SR) checkpoints? #3

Open escorciav opened 3 months ago

escorciav commented 3 months ago

Congratulations on your CVPR acceptance! 🎉

Could you share any checkpoints, along with the inference and/or training code, to reproduce the super-resolution results?

Thank you for making your work accessible and reproducible!

MKFMIKU commented 3 months ago

Hi, escorciav. Thanks for your attention to our work.

Sorry that we cannot share any checkpoint according to our internal policy. However, you should be able to reproduce the results according to the released code.

Moreover, you can find our 4-step super-resolution results / LR images / HR images at https://drive.google.com/file/d/1aY1xDk-l7-ErbTZi5VL7XNg71a9HZSEs/view?usp=sharing

escorciav commented 3 months ago

Gotcha. Do you mind sharing more details about the training & inference recipe? I'm not familiar with the workflow/pipeline.

I was expecting to grab your model & develop stuff to deploy/run it on mobile :smiley: Mobile target: Qualcomm Snapdragon 8 Gen >= 2 chipset