Dimensions of processed SHREC2017

VIPL-SLP / pointlstm-gesture-recognition-pytorch

This repo holds the codes of paper: An Efficient PointLSTM for Point Clouds Based Gesture Recognition (CVPR 2020).

https://openaccess.thecvf.com/content_CVPR_2020/html/Min_An_Efficient_PointLSTM_for_Point_Clouds_Based_Gesture_Recognition_CVPR_2020_paper.html

Apache License 2.0

122 stars 20 forks source link

Dimensions of processed SHREC2017 #20

Open czhou005 opened 2 years ago

czhou005 commented 2 years ago

Hi there,

After processing the SHREC2017 by the respective process.py, I found each video has been converted to pts_label.npy in [32, 256, 8]. The 32 refers to sampled video frame numbers, whereas each frame has 256 points sampled. Could you elaborate on 8? Does this mean channel or something pertinent?

Many thanks for considering this issue.

ycmin95 commented 2 years ago

Hi, thanks for your attention. As shown in the L35-36 of the process code, the [0,4) correspond to the x,y,z,t; and the [4,7) correspond to the transformed x, y, z in the world coordinate system, and 7 is still the timestep. The transformation process can be found here. We provide this code to evaluate the spatial information under different coordinate systems.