Zongsheng Yue comments

Results 17 comments of


                                            Zongsheng Yue

add pipeline of DifFace

@patrickvonplaten Hi, My code depends on the library of opencv, which causes some failed tests. Is it possible to include opencv in diffusers since it is a very common tool...

@williamberman Our project depends on another deep image super-resolution model [Real-ESRGAN](https://github.com/xinntao/Real-ESRGAN) and face detection model [RetinaFace](https://github.com/xinntao/facexlib). They are not necessary for diffusers but specified for our project. I don't know...

self.loss_fun in trainer.py is not defined.

We have corrected this bug and the configuration file. Please re-download our repository. Thanks. @bwhwang

cv2.error: OpenCV(4.8.0) :-1: error: (-5:Bad argument) in function 'imencode' > Overload resolution failed: > - Can't parse 'params'. Sequence item with index 1 has a wrong type > - Can't parse 'params'. Sequence item with index 1 has a wrong type

It seems that opencv failed to read your image. Please have a try read your image with opencv.

It's not convenient to change the cuda device when running inferrence script.

Please use CUDA_VISIBLE_DEVICES directly.

module diffusers has no attribute DifFacePipeline

Sorry. We haven't added DifFace to diffusers.

Model weight preservation in the Train diffused estimator (SwinIR) stage.

You should load the key of "state_dict". I have extracted this key for the released model weight.

Does adjustment of lr work?

No, the learning rate is fixed as 1e-4 following the setting in DDPM.

Does adjustment of lr work?

We have not trained the model with 1024 x 1024, which requires huge computation resources.

Could you release your metric calculation script please?

For PSNR and LPIPS: https://github.com/chaofengc/IQA-PyTorch For LPIPS (VGG): https://github.com/richzhang/PerceptualSimilarity For IDS: see the script of VQFR. For FID: https://github.com/mseitzer/pytorch-fid As for as I know, for the metric of FID, VQFR...