Data Preparation

[tommyshelby4]

I get an error when trying to load the data for inference.


When I debug it, I see it's totally normal as the .npz files do not contain any 'dist' or 'nn_pose' key entries. I took a look at the data preparation scripts and I cannot see in which way they could be solving this issue. Can anybody who has managed to run the code properly help me go all the way?

[HospitableHost]

you need to take a look at the file 'prepare_data.py' and 'prepare_traindata.py'.

[tommyshelby4]

I have taken a look but I keep getting errors like that:

Do you have any idea?

[HospitableHost]

I have taken a look but I keep getting errors like that: Do you have any idea?

why ( _, 21 , 3 )? Is't ( _, 21 , 4 )?

[tommyshelby4]

Why should it be 4?

[garvita-tiwari]

I have taken a look but I keep getting errors like that: Do you have any idea?

why ( _, 21 , 3 )? Is't ( _, 21 , 4 )?

It is 3 here, because we are loading AMASS poses, in axis angle representation(not quaternion).

[garvita-tiwari]

I have taken a look but I keep getting errors like that: Do you have any idea?

This looks like that

1. either you are using a different no. of joints
2. or you are loading a data, which was saved differently.

Did you run sample_poses.py before running prepare_data.py??
Please share the shape of pose_seq after this line:
pose_seq = np.load(os.path.join(ds_dir, seq))['pose_body']

[tommyshelby4]

Thanks for your reply. I haven't run sample_poses.py because I get this error while the PoseNDF env is activated

[garvita-tiwari]

Thanks for your reply. I haven't run sample_poses.py because I get this error while the PoseNDF env is activated

This error has nothing to do with the environment. There is file data/data_splits.py which basically contains the AMASS train/val/test split. "from data.data_splits import amass_splits" access this.

I guess this is because of path. Try changing "from data.data_splits import amass_splits" to "from data_splits import amass_splits"