search

AnweshCR7 / RhythmNet

End-to-end Heart Rate Estimation from Face via Spatial-temporal Representation. A replication of the RhythmNet model.
https://arxiv.org/abs/1910.11515
MIT License
58 stars 13 forks source link

Pearson's r #4

Open GKFXCode opened 3 years ago

GKFXCode commented 3 years ago

Hi, have you calculate Pearson's r when eval? I tried to reproduce the results reported on the paper but find r is very low when eval.

AnweshCR7 commented 3 years ago

I haven't calculated the Pearson r for this model; in my opinion, this kind of correlation should apply in the case of a signal and not point values. However, the metrics presented in the paper are derived using a 5-fold evaluation. If you have a considerable difference in the values, maybe you should do the 5-fold cross-validation.