8000 About Face Swapping Results and Applicability of crop_and_mask · Issue #18 · Sanoojan/REFace · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

About Face Swapping Results and Applicability of crop_and_mask #18

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
frujji opened this issue Jun 4, 2025 · 0 comments
Open

About Face Swapping Results and Applicability of crop_and_mask #18

frujji opened this issue Jun 4, 2025 · 0 comments

Comments

@frujji
Copy link
frujji commented Jun 4, 2025 •
edited
Loading

First, thank you for providing the complete code. We're highly interested in your face swapping work and have been attempting to reproduce it. During our implementation process, we encountered the following issues and would greatly appreciate your guidance:

1、Face Swapping Results
For the example images provided in the source code, I used crop_and_mask.py to perform alignment, cropping, and mask generation (with mask_real_path set to None). Then input the aligned images and their corresponding masks into inference_test_bench.py to generate face-swapped results. However, the output differs from the results shown in your paper. Could you clarify if I missed any important steps or if there are errors in my workflow? As shown in the figure, the second and fourth columns display the visualized mask results after applying crop_and_mask.

Image

2、Applicability of crop_and_mask
Is crop_and_mask universally applicable to all face datasets without provided masks?
Since the above results did not meet expectations, I further tested the crop_and_mask pipeline by generating masks for images from CelebA-HQ. The resulting masks were inconsistent with those produced by process_CelebA_mask. Do you have any comments on this discrepancy? The figure below illustrates the differences between the masks generated by the two methods. Note that I enhanced the color contrast for better visualization, but the actual masks were generated strictly following the source code's workflow.

Image
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@frujji
0