titu1994/MLSTM-FCN

Confusion for The Number of variables in Action 3D dataset

YaoDongAn opened this issue · 1 comments

Hi ,titu~ @titu1994

I have read your papers about time series classification and your great work sparked my interest!
So I want figure out the detail of this master work.

I noticed that the number of variables in Action 3D dataset is 570. And I am confused how this number is calculated.
As far as I know, The data structure of Action 3D dataset is 20*4 (in single time frame) . 20 means there are 20 skeleton points, 4 means each skeleton point is described by 4 dimension:(x,y,d,c). x,y represent the position of skeleton point in the picture,and d represent depth,c represents the score confidence.

Could you nicely tell me how you extract 570 variables form the inital datasets?
It confused me for a long time, hope you can help.
Thanks anyway!

I figure it out that 570 = 190*3,which represent 190 pairwise relative keypoints positions in 20 keypoints。and 3 means location x,y,z。