search

ubicomplab / rPPG-Toolbox

rPPG-Toolbox: Deep Remote PPG Toolbox (NeurIPS 2023)
https://arxiv.org/abs/2210.00716
Other
504 stars 129 forks source link

About data preprocessing #280

Closed 408550969 closed 5 months ago

408550969 commented 5 months ago

Hello, I am not familiar with RPPG. I would like to confirm if my data preprocessing is correct. Firstly, I collected PPG signals. For example, I collected 10 videos, each lasting 1 minute. Then I divided the video into images and matched the PPG signal with each image. Then I normalized the PPG signal of each video segment with a mean of 0 and a standard deviation of 1. Finally, conduct training. What I want to ask is whether the normalization operation is performed separately for each video segment (i.e. 10 videos are normalized 10 times each), or for all videos (i.e. all training sets are normalized only once)?

408550969 commented 5 months ago

I noticed that in the UBFC-rPPG_UBFC-rPPG-PURE EFFICIENTPHYS.yaml file, DATA-TYPE: ['Standardized '] means that normalization will be performed when reading data, and the label of UBFC-rPPG itself has already been normalized, which is equivalent to two rounds of normalization. Will this affect the results?