Output value

[Q-Y-Yang]

Hey,

Thanks a lot for the repository.

I unpacked some pkl file from skeletons_pkl/D9H-eRIROA to see your processed results.
I think the "Open pose" array is the pred_joints_img in the pred_output_lists, which stores the 3D joints coordinates.

For example:
{'open_pose': array([[ 574.05596924, 179.28507996, -127.43224335],
[ 534.85345459, 289.98730469, -48.78568268],
[ 440.2225647 , 323.91088867, -64.36065674],
[ 377.38534546, 479.19961548, -67.28412628],
[ 457.2668457 , 531.45391846, -187.31272888],
[ 655.25250244, 326.978302 , -58.89009476],
[ 695.20507812, 482.97067261, -36.92470932],
[ 632.99291992, 490.90283203, -179.33460999],
[ 543.42022705, 579.68896484, 8.63186359],
[ 505.92532349, 644.42492676, 27.28130531],
[ 462.63250732, 849.27270508, -26.45468521],
[ 487.91247559, 990.62567139, 172.10594177],
[ 578.80175781, 636.48510742, 28.43116951],
[ 606.41992188, 850.32006836, -44.75795364],
[ 616.98632812, 960.39355469, 170.74185181],
[ 550.43334961, 156.90029907, -106.82781982],
[ 589.61871338, 162.80293274, -95.0479126 ],
[ 508.64923096, 179.9957428 , -62.24282455],
[ 593.3503418 , 188.85913086, -38.49153137],
[ 624.86157227, 1070.73754883, 112.74629211],
[ 666.54736328, 1049.30419922, 134.95561218],
[ 635.23919678, 944.24407959, 213.73262024],
[ 473.74057007, 1092.18286133, 110.29068756],
[ 434.71447754, 1070.87841797, 145.2804718 ],
[ 494.05203247, 982.76074219, 227.28088379]], dtype=float32)

I doubt that the third column is the depth information, cuz there are always positive and negative values. Is it right?
In the docs of FrankMocap it only said 'pred_joints_img': #3D joints where X,Y are aligned to input image, and did not mention depth.

Besides, I also obtained some outputs of FrankMocap with a webcam at lab. The third column of pred_joints_img also contains both positive and negative values.

In the body_mocap_api.py, there is "pred_joints_3D", which is the 3D joints positions.

Thanks!

[cotton-ahn]

• Yes, the third column is the depth information :)

[Q-Y-Yang]

• Yes, the third column is the depth information :)

Thanks.

Intuitively I do not understand why there are both positive and negative values. Isn't the origin in the camera?
I googled, but did not find useful information...