search

ubicomplab / rPPG-Toolbox

rPPG-Toolbox: Deep Remote PPG Toolbox (NeurIPS 2023)
https://arxiv.org/abs/2210.00716
Other
503 stars 127 forks source link

Use of 'PURE-PURE-iBVP-iBVPNet' #330

Open creaycoder opened 1 week ago

creaycoder commented 1 week ago

I'm using the iBVPNet model. Specifically, I use 'PURE-PURE-iBVP-iBVPNet.yaml' to use the iBVPNet model. Due to the collective size of the data, I only used a portion of the two data sets described. I ran the model training and validation all right, but I ended up with the following error, and the folder 'saved_test_outputs' was not found, and the generated folder 'saved_test_outputs' was empty. I would like to ask you, what is the reason? Have you encountered it? 6a61a5e3964e17d5a0b13cf2bd44124 8af4a0b19b8439c8239869936e72504

yahskapar commented 1 week ago

Hi @creaycoder,

I think this is because of your sample size (in this case, number of videos in the test dataset) is too small, and therefore the covariance matrix for the Pearson correlation calculation isn't well-defined. Can you try increasing the test dataset to perhaps add a few more videos (perhaps two more on top of the existing two)?

Alternatively, in the TEST section of your config file, you can just remove the Pearson metric. That should avoid this error and allow you to see the additional test-time outputs.