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

Some excellent demo #31

Closed
lucasjinreal opened this issue Apr 25, 2023 · 12 comments
Closed

Some excellent demo #31

lucasjinreal opened this issue Apr 25, 2023 · 12 comments

Comments

@lucasjinreal
Copy link

Hi, just post some FBX (not rending, in real 3D) demos here, the result is impressive:

Clip_len 24

2023-04-25-14-47-18 00_00_00-00_00_30

Clip_len 48

2023-04-25-15-08-26 00_00_00-00_00_30

the video I tested is a very challenge one, still get some nice result!

Just still have one issue, the poses might blink in middle of frames. Do u got any thoughts? What's more, what's the best clip len here for realtime applications? (we can't using too big clip len here in realtime)

@Walter0807
Copy link
Owner

Walter0807 commented Apr 25, 2023

Very cool! Glad to know you've got good results.

  1. You can check the 2D pose estimation results first. For example, are there some severe inconsistencies due to occlusion?
  2. Jittering might also happen due to clip divisions. It might get improved with some post-processing steps (e.g. moving average with sliding window) to ensemble the predictions.
  3. Yes, you can try smaller clip lengths and see how it works. Some smoothing techniques (as mentioned above) might be useful in that case.

@Walter0807
Copy link
Owner

Btw, would you mind sharing the scripts or the pipeline for exporting FBX animations? I believe it would be useful for many people. Thank you very much!

@lucasjinreal
Copy link
Author

Sure, the script I use is from ROMP, I just save motionbert result into pickle file and same way load in ROMP.

@Walter0807 Walter0807 pinned this issue Apr 27, 2023
@zhangjinyangnwpu
Copy link

zhangjinyangnwpu commented May 10, 2023

We can use the script here[https://github.com/Arthur151/ROMP/blob/master/simple_romp/tools/convert2fbx.py] to export fbx

@dddyyyplus
Copy link

dddyyyplus commented May 23, 2023

Sure, the script I use is from ROMP, I just save motionbert result into pickle file and same way load in ROMP.

@lucasjinreal Hi, can I ask which motionbert result you use here, 3d mesh or 3d pose estimation results? Thank you!

@TnoobT
Copy link

TnoobT commented Jun 6, 2023

@lucasjinreal Hello, could you please share the script for converting the output of motionbert to the input data of convert2fbx.py. Thank you very much

@August1996
Copy link

@lucasjinreal 请问怎么用convert2fbx来转换motionBERT输出的结果。我输入使用3d pose的npy文件,但是转换失败了
image

@August1996
Copy link

Sure, the script I use is from ROMP, I just save motionbert result into pickle file and same way load in ROMP.当然,我使用的脚本来自 ROMP,我只是将 motionbert 结果保存到泡菜文件中,并以同样的方式加载到 ROMP 中。

这里需要修改motionBERT或者ROMP的脚本代码吗?

@DenisKochetov
Copy link

Sure, the script I use is from ROMP, I just save motionbert result into pickle file and same way load in ROMP.当然,我使用的脚本来自 ROMP,我只是将 motionbert 结果保存到泡菜文件中,并以同样的方式加载到 ROMP 中。

这里需要修改motionBERT或者ROMP的脚本代码吗?

Hi, did you find out how to convert it to fbx?

@HoangNguyenHuu
Copy link

HoangNguyenHuu commented Oct 28, 2023

Sure, the script I use is from ROMP, I just save motionbert result into pickle file and same way load in ROMP.

Thanks for this great work!
@lucasjinreal, @Walter0807. Hi, I have questions. ROMP export to SMPL 24 joints, MotionBert export to H36M 17 joints . How do you save MotionBERT output into a pickle file like ROMP output?
In ROMP, results have "cam_trans" data, how to get it in MotionBERT.
Thank you very much!

@August1996
Copy link

cam_trans

You can reference this link

@DenisKochetov
Copy link

Thank you, recently found same code in the original video2bvh

cam_trans

You can reference this link

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

8 participants