kebii/tapmo

tapmo's Issues

Training code?

Hi,

Thanks for this amazing work! I was wondering when you are planning on releasing the training code.

ValueError: not enough values to unpack

I directly run the first command in shape_diffusion and got the following error

Could someone please check that?

Missing `Mean_30p_6d.npy` and `Std_30p_6d.npy`

python3 -m sample.generate_handle_motion --model_path ../weights/diffusion_model_latest.pt --arch trans_dec --emb_trans_dec False --dataset t6d_mixrig --char_feature_path ../demo/shape_features/001.npy --save_path ../demo/motion/motion_3 --text_prompt "walk forward and turn right."
Loading model args from [../weights/args.json]
Warning: was not able to load [char_feature_path], using default value [../demo/shape_features/001.npy] instead.
Warning: was not able to load [save_path], using default value [../demo/motion/motion_3] instead.
Loading dataset...
get dateset: t6d_mixrig
Loading dataset t6d_mixrig ...
Traceback (most recent call last):
  File "/home/zjp/anaconda3/envs/sfpt/lib/python3.9/runpy.py", line 197, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/zjp/anaconda3/envs/sfpt/lib/python3.9/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/home/zjp/TapMo/shape_diffusion/sample/generate_handle_motion.py", line 205, in <module>
    main()
  File "/home/zjp/TapMo/shape_diffusion/sample/generate_handle_motion.py", line 66, in main
    data = load_dataset(args, max_frames, n_frames)
  File "/home/zjp/TapMo/shape_diffusion/sample/generate_handle_motion.py", line 195, in load_dataset
    data = get_dataset_loader(name=args.dataset,
  File "/home/zjp/TapMo/shape_diffusion/data_loaders/get_data.py", line 50, in get_dataset_loader
    dataset = get_dataset(name, num_frames, split, hml_mode)
  File "/home/zjp/TapMo/shape_diffusion/data_loaders/get_data.py", line 43, in get_dataset
    dataset = DATA(split=split, num_frames=num_frames, mode=hml_mode)
  File "/home/zjp/TapMo/shape_diffusion/data_loaders/humanml/data/dataset_t6d_mixrig.py", line 874, in __init__
    self.mean = np.load(pjoin(datapath, 'Mean_30p_6d.npy'))[None,]
  File "/home/zjp/anaconda3/envs/sfpt/lib/python3.9/site-packages/numpy/lib/npyio.py", line 427, in load
    fid = stack.enter_context(open(os_fspath(file), "rb"))
FileNotFoundError: [Errno 2] No such file or directory: '../datasets/HumanML3D/Mean_30p_6d.npy'

Thanks for sharing code!
I download the datasets.zip from google drive but missing the files.

When will the code be open sourced?

Fantastic work! Will the code be open sourced? and when?

Question about the ground truth motion for training

Do you mean you use ICP to transfer the motion of SMPL to other characters?

Missing args.json

python3 -m sample.generate_handle_motion --model_path ../weights/diffusion_model_latest.pt --arch trans_dec --emb_trans_dec False --dataset t6d_mixrig --char_feature_path ../demo/shape_features/001.npy --save_path ../demo/motion/motion_3 --text_prompt "walk forward and turn right."
Loading model args from [../weights/args.json]
Traceback (most recent call last):
  File "/home/zjp/anaconda3/envs/sfpt/lib/python3.9/runpy.py", line 197, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/zjp/anaconda3/envs/sfpt/lib/python3.9/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/home/zjp/TapMo/shape_diffusion/sample/generate_handle_motion.py", line 205, in <module>
    main()
  File "/home/zjp/TapMo/shape_diffusion/sample/generate_handle_motion.py", line 19, in main
    args = generate_args()
  File "/home/zjp/TapMo/shape_diffusion/utils/parser_util_tapmo.py", line 226, in generate_args
    return parse_and_load_from_model(parser)
  File "/home/zjp/TapMo/shape_diffusion/utils/parser_util_tapmo.py", line 22, in parse_and_load_from_model
    assert os.path.exists(args_path), 'Arguments json file was not found!'
AssertionError: Arguments json file was not found!

Thanks for sharing the inference code!
I download the weights.zip from google drive and unzip it into the folder but missing args.json file to get the arguments, could you provide it?

Customized model

Hi,

I checked the demo results' obj and it looks great. Tho I do have some customization request:

1. How to run TapMo on a customized model? i.e. given an customized .obj file, not from the demo. I notice that you use the shape_features/001.npy at step 3, not a standard .obj file.

2. How to have more than currently 120 .obj outputs?

Appreciate your help!

Missing /datasets/HumanML3D/HumanML3D/test.txt'

Hello,
I'm trying to run the code, seems like there are still files missing.

Could you please upload it? BTW, if possible, could you please double check by running the two inference command on your end to ensure all the needed files are uploaded? Huge thanks in advance.

How to get skeleton-driven mesh?

How do you obtain the skeleton-driven mesh mentioned in the paper? The article doesn't seem to explain this in great detail.