RhythmNet: End-to-end Heart Rate Estimation from Face via Spatial-temporal Representation

A reproduction of the RhythmNet model. Paper link

Dataset:

VIPL-HR dataset

Experiments

Shared parameters:

batch size: 32
Dataset: VIPL
Model: RhythmNet
initial learning rate: 1e-3
epochs: 50
window_size = 300 frames with stride of 0.5 seconds

Dataset-split: 5 fold validation

Experiment for 1-Fold without GRU layer

Set	Loss	MAE (bpm)	RMSE (bpm)
Training	3.096	1.817	2.834
Eval	15.91	9.255	11.787

Experiment for 1-Fold with GRU layer

Set	Loss	MAE (bpm)	RMSE (bpm)
Training	3.925	2.423	4.16
Eval	14.25	13.992	17.019