search

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
117 stars 19 forks source link

May I ask difference between SHREC data and NVGesture data pre-processing? #6

Closed doyeon16 closed 3 years ago

doyeon16 commented 3 years ago

Hi, @Blueprintf

Thank you for sharing your code.

I have a question about the code.

Unlike pre-processing code of SHREC data, NVGesture subtracts the value (= pts[:, 2]) from 255 when it preprocesses the data (in uvd2xyz_nvidia funtion in utils.py ). May I ask why? Is there the difference between when you did it and when you didn't?

Thanks for reading

ycmin95 commented 3 years ago

@doyeon16 If you visualize the NVGesture data, you will find that the closer pixels have larger depth values, which is different from the usual setting. We subtract the depth value (= pts[:, 2]) from 255 to do pose estimation in the initial attempts and keep this setting. Hope this can help you~

doyeon16 commented 3 years ago

@Blueprintf Oh I see. Thanks for the quick reply.