cuixing100876

I replace the code with yours, but I got the output as this

2

I replace the code with yours, but I got the output as this _Originally posted by @phhandong in https://github.com/diyiiyiii/StyTR-2/issues/10#issuecomment-1225680160_

It seems that draggan can't handle natural images in the dragbench dataset, because there is no way to inversion these images accurately. How did you conduct the comparison with draggan

How to compute Fidelity (1-LPIPS)? It seems not lpips score or clip score.

Question about the learnable queries in DiffEditor

1

Hi. It seems that the learnable query in the code is randomly initialized. (src/unet/attention_processor.py class Resampler line 321). Besides, there does not appear to be any other initialization of the...

Could you release the test set or share the corresponding code to obtain your test set?