search

Xinyu-Yi / TransPose

A real-time motion capture system that estimates poses and global translations using only 6 inertial measurement units
https://xinyu-yi.github.io/TransPose/
GNU General Public License v3.0
392 stars 77 forks source link

Can not reproduce test result in TotalCapture #47

Closed iamyouno closed 8 months ago

iamyouno commented 2 years ago

Hi, I'm trying to reproduce test result using DIP-IMU and TotalCapture datasets.

When run preprocess.py and evaluate.py, It worked well in DIP-IMU, but not in TotalCapture.

Is anyone who reproduce test result with TotalCapture dataset?

Should I change code or some data?

yul97 commented 1 year ago

Hi, I'm trying to reproduce test result using DIP-IMU and TotalCapture datasets.

When run preprocess.py and evaluate.py, It worked well in DIP-IMU, but not in TotalCapture.

Is anyone who reproduce test result with TotalCapture dataset?

Should I change code or some data?

Me too. Have you found the problem?

Xinyu-Yi commented 1 year ago

It should reproduce the TotalCapture results in the paper without changing any code (except for some paths in the config). Are you using the ground truths provided by the DIP authors?

yul97 commented 1 year ago

It should reproduce the TotalCapture results in the paper without changing any code (except for some paths in the config). Are you using the ground truths provided by the DIP authors?

I checked it and still haven't solved this problem. Could you tell me whether the ground truths provided by the DIP authors is the dataset shown in the figure？ image

Xinyu-Yi commented 1 year ago

I remember that I sent an email to DIP authors and got the data. I'm not sure whether your data is right or not. I think you can check the results visually.

syguan96 commented 1 year ago

Hi guys, have you solved this problem? My reproduced results are: ============== offline ================ SIP Error (deg): 62.50 (+/- 22.71) Angular Error (deg): 50.06 (+/- 17.78) Positional Error (cm): 28.58 (+/- 10.20) Mesh Error (cm): 31.46 (+/- 11.38) Jitter Error (100m/s^3): 1.47 (+/- 2.85)

Xinyu-Yi commented 1 year ago

Hi guys, have you solved this problem? My reproduced results are: ============== offline ================ SIP Error (deg): 62.50 (+/- 22.71) Angular Error (deg): 50.06 (+/- 17.78) Positional Error (cm): 28.58 (+/- 10.20) Mesh Error (cm): 31.46 (+/- 11.38) Jitter Error (100m/s^3): 1.47 (+/- 2.85)

Please send me your saved test results. I will check this.

syguan96 commented 1 year ago

Hi Xinyu @Xinyu-Yi , I download SMPL annotations of TotalCapture from the official website.

The downloaded SMPL annotation and processed file are uploaded. You can download it from https://cowtransfer.com/s/8388217ba9e340 .

I uncommented evaluate_pose(paths.totalcapture_dir), and then run python evalaute.py. You can check it.

scholars-Lwp commented 1 year ago

@syguan96
I have the same problem as you. Have you solved it? If so, can you tell me what the problem is?

scholars-Lwp commented 1 year ago

I have the same problem as you. Have you solved it? If so, can you tell me what the problem is?

Xinyu-Yi commented 1 year ago

Sorry for the late response. I have uploaded the old version of the SMPL pose (obtained from DIP authors) for the TotalCapture dataset. See https://github.com/Xinyu-Yi/PIP/issues/34

bryanbocao commented 1 month ago

Hi guys, have you solved this problem? My reproduced results are: ============== offline ================ SIP Error (deg): 62.50 (+/- 22.71) Angular Error (deg): 50.06 (+/- 17.78) Positional Error (cm): 28.58 (+/- 10.20) Mesh Error (cm): 31.46 (+/- 11.38) Jitter Error (100m/s^3): 1.47 (+/- 2.85)

I got the same results as well. May I know if you solved the problem? Thank you! @syguan96