zh-plus/video-to-pose3D

Normalize/unNormalize Data

mmaGroupGitHub opened this issue · 1 comments

mmaGroupGitHub commented

Could you explain and point to me in your code where you are perform normalization and unNormalization of the data.

If I just try to plot the 3D output the values are all less than 1 just trying to understand where you perform unNormalization and also what sort of preprocessing you apply to the two dimensional keypoints.

zh-plus commented

The 2D normalization:

video-to-pose3D/videopose.py

Line 79 in e271582

keypoints = normalize_screen_coordinates(keypoints[..., :2], w=1000, h=1002)

The 3D unNormalization may be performed here:

video-to-pose3D/common/utils.py

Line 191 in e271582

# Undo flipping and take average with non-flipped version

The 3D Normalization in the function model_pos:

video-to-pose3D/common/utils.py

Line 189 in e271582

predicted_3d_pos = model_pos(inputs_2d)

I'm not sure what "Normalization" you said referred to. For more related information, you may search Normalization in the original videopose3D repo.

Sorry for the late response.