Skip to content
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.

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

Inconsistency in DensePose Visualisation Across Consecutive Frames #19

Open
yihong1120 opened this issue Dec 11, 2023 · 1 comment
Open

Comments

@yihong1120
Copy link

Dear Vid2DensePose Maintainers,

I hope this message finds you well. I am reaching out to report an issue I've encountered while utilising the Vid2DensePose tool for a project aimed at enhancing animation sequences through detailed human pose estimation.

Issue Description:
Upon generating the "Part Index" visualisations for a sequence of video frames, I observed an inconsistency in the visual output across consecutive frames. Despite the input video depicting minimal movement between frames, the DensePose visualisation appears to fluctuate, leading to a jittery and discontinuous representation of the pose data.

This inconsistency poses a challenge in achieving the desired temporal coherence for human image animation, particularly when integrating with MagicAnimate. The expected outcome is a smooth transition of pose visualisations that accurately reflect the movement within the video.

Steps to Reproduce:

  1. Follow the installation and usage guide as per the repository's README.md.
  2. Run the main.py script on a video with subtle movements.
  3. Observe the output video for any discrepancies in the DensePose visualisation across frames.

Potential Impact:
The issue has the potential to affect users who rely on the tool for high-fidelity animation projects, where precision and consistency in human pose estimation are paramount.

Request:
I kindly request your assistance in addressing this matter. Any insights into potential causes or suggestions for mitigating this inconsistency would be greatly appreciated. Additionally, if there are any recommended practices or parameters that could enhance the stability of the visual output, I would be eager to learn and implement them.

Acknowledgments:
I would like to extend my gratitude for your efforts in developing such a valuable tool for the community. Your work significantly contributes to the advancement of human pose estimation applications.

Thank you for your time and support. I look forward to your response and any guidance you can provide.

Best regards,
yihong1120

@samsara-ku
Copy link

Did you solve this problem? In my case, it is not still working well like below:

image
image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants