DiffFace icon indicating copy to clipboard operation
DiffFace copied to clipboard

DiffFace: Diffusion-based Face Swapping with Facial Guidance

Results 16 DiffFace issues
Sort by recently updated
recently updated
newest added

> Traceback (most recent call last): > File "/data/docker/yerang/DiffFace/main.py", line 6, in > image_editor = ImageEditor(args) > File "/data/docker/yerang/DiffFace/optimization/image_editor.py", line 123, in __init__ > self.fa = face_alignment.FaceAlignment(face_alignment.LandmarksType._2D, flip_input=False) > File...

Traceback (most recent call last): File "main.py", line 7, in image_editor.edit_image_by_prompt() File "/root/DiffFace/optimization/image_editor.py", line 393, in edit_image_by_prompt visualization_path = visualization_path.with_stem(f"{visualization_path.stem}_i_{iteration_number}_b_{0}") AttributeError: 'PosixPath' object has no attribute 'with_stem'

it seems that the model didn't work well when source and target face have different mask orientation. ![output_i_0_b_0](https://github.com/hxngiee/DiffFace/assets/37773231/2016e495-2472-4d65-96c9-d814cdf2477b)

I got this error By the way I have rtx 3090 will this work with me ![image](https://github.com/hxngiee/DiffFace/assets/107565623/20812fcc-b220-4839-91df-05e8fbef567a)

Does the current released code version support DDIM sampling? I have tried to tag the "--ddim", but got weird results

Hello, author, thank you for sharing. I have a question: Do you have a specific method for aligning the faces of the source image and the target image? Looking forward...

Is your model brings pretraied DDPM(trained with FFHQ) and finetune with loss? In your code, it seems like you finetune the model per each image using Model.pt, but there is...