xiexh20/chore

Official implementation for the paper CHORE, ECCV'22

chore's Issues

No contact

Hello,

Thank you for sharing your great work! I am very interested in your work and try to run the demo. However, I found that the contact cannot be detected. The log and results are as follows. Is this a parameter setting problem or something else? Thank you.

chore_log.txt

How to test BEHAVE test set?

@xiexh20 Thank you very much for your great work!

The python script don't work so that I can't test CHORE on BEHAVE test set as follows.

From your github,

Test on BEHAVE test set
After downloading the pretrained model, you can test one sequence from behave data with:

python recon/recon_fit_behave.py chore-release --save_name chore-release -s [path to one sequence]

I just try
python recon/recon_fit_behave.py chore-release --save_name chore-release -s /data2/BEHAVE_data/test/Date03_Sub03_backpack_back (This is my folder pass)

Then I have next results. Would you show me how to test CHORE on BEHAVE test set? Thank you very much!

configs loaded from config/chore-release.json
In total 0 test examples
Loaded checkpoint from: /home/work/CHORE/recon/../experiments/chore-release/checkpoints/checkpoint_38h:1m:33s_136893.18501019478.tar
Date03_Sub03_backpack_back: : 0it [00:00, ?it/s]
all done

Invalid device ordinal

Hi.
Thanks for the great work. I'm trying to test your code on my own data. Unfortunately, I met an error when I run the demo.py, as shown below.

My environment is nvidia RTX 3090ti with cuda 11.3 and pytorch 1.10.
I am wondering whether it is a cuda version problem and if it is, can this code run on RTX 30xx device (Since RTX 3090ti could only fit cuda version 11+)?

Thank you.

Installation for detectron2 fails when using pytorch 1.6.0+cu101

Hi, thanks for the great work. I am trying to run the demo but am failing since the installation process you have provided doesn't work.
When running pip install -r requirements.txt with torch version 1.6.0, I face this error:

'''''''''''''''''''''''''''''''''''''Error'''''''''''''''''''''''''''''''''''''
Collecting git+https://github.com/facebookresearch/detectron2.git
Cloning https://github.com/facebookresearch/detectron2.git to /tmp/pip-req-build-ri2eeos8
Running command git clone --filter=blob:none --quiet https://github.com/facebookresearch/detectron2.git /tmp/pip-req-build-ri2eeos8
Resolved https://github.com/facebookresearch/detectron2.git to commit 3c7bb714795edc7a96c9a1a6dd83663ecd293e36
Preparing metadata (setup.py) ... error
error: subprocess-exited-with-error

× python setup.py egg_info did not run successfully.
│ exit code: 1
╰─> [6 lines of output]
Traceback (most recent call last):
File "", line 36, in
File "", line 34, in
File "/tmp/pip-req-build-ri2eeos8/setup.py", line 14, in
assert torch_ver >= [1, 8], "Requires PyTorch >= 1.8"
AssertionError: Requires PyTorch >= 1.8
[end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed

× Encountered error while generating package metadata.
╰─> See above for output.

note: This is an issue with the package mentioned above, not pip.
'''''''''''''''''''''''''''''''''''''Error'''''''''''''''''''''''''''''''''''''

How exactly did you install Detectron2? Can you provide the exact installation procedure?

There is something wrong in the GT

Thanks for your interesting work.
I have tried to test on the BEHAVE dataset with CHORE.
But i found there is some thing wrong in the dataset.
For some frames, the joint values of k1.color.json are all 0.
For example, the frame of Date03/Date03_Sub03_backpack_back/t0029.000/k1.color.json.
Could you check the GT and give me some advice?

Doubts about training

Hello, I try to train on the behave dataset, but I always get an error "cuda out of memory". I would like to ask what GPU you used for training? In addition, I would like to ask that you mentioned that you also used COCO and NTU-RGBD datasets for training in your paper. Can you provide relevant training codes and NTU-RGBD datasets?

Which SMPL+H Body Model to use & where to extract?

In your installation guide it says

We use the SMPL-H body model, please prepare it from the official website and modify SMPL_MODEL_ROOT in PATHS.yml accordingly.

However, in the Download section there are multiple archives to download (Models & Code, Extended SMPL+H Model, Scans, Registrations...).

1. Which one do I have to use?
2. Does it matter where I extract the archive?
3. Furthermore, which directory do I have to enter in the SMPL_MODEL_ROOT variable (I guess the root directory of the archive)?

Thank you for your support!

About evaluation

Can't generate processed data

Reconstruction time

Hi,

Thanks for releasing this code. It is truly a very exciting work!

I was wondering how much time it takes on average to make one estimation with your method? I run some examples on my PC, it takes me around 03:30 minutes for one image reconstruction.

Is this the same time it takes you? Any advice on how to run the optimization faster?

Thanks in advance : )