Predict wrong joints

[tiffany0117]

Hello everyone,

I'm trying to rebuild the result of this code and follow the introduction that provide on this website, however, I got the wrong predicted joints, which are not the 16 joints we want to show the skeleton.
Could anyone help me change the code to train the model with the joints we want?

1. Operating system : windows 10 home
2. Tensorflow version : 2.3.0
3. Python version : 3.8.8

[una-dinosauria]

Could you please elaborate? How are they wrong?

[tiffany0117]

When I run the code in the first time, I got nothing on visualize in prediction, so I checked the variable, poses3d, I found that except for hip joint, all joints list below are all Nan, however there is value in the others joints that not list below.

H36M_NAMES[0] = 'Hip'
H36M_NAMES[1] = 'RHip'
H36M_NAMES[2] = 'RKnee'
H36M_NAMES[3] = 'RFoot'
H36M_NAMES[6] = 'LHip'
H36M_NAMES[7] = 'LKnee'
H36M_NAMES[8] = 'LFoot'
H36M_NAMES[12] = 'Spine'
H36M_NAMES[13] = 'Thorax'
H36M_NAMES[14] = 'Neck/Nose'
H36M_NAMES[15] = 'Head'
H36M_NAMES[17] = 'LShoulder'
H36M_NAMES[18] = 'LElbow'
H36M_NAMES[19] = 'LWrist'
H36M_NAMES[25] = 'RShoulder'
H36M_NAMES[26] = 'RElbow'
H36M_NAMES[27] = 'RWrist'

And here are the value of the poses3d :

[una-dinosauria]

Please give TF 1.2 a try: #197 (comment)

[tiffany0117]

THK!!!