Micro Expression Recognition via Dual-Stream Spatiotemporal Attention Network

Table 2

The influence of different modules on the network.

Method	CASME II
	Accuracy	F1-score
Basic model	0.655	0.637
Basic model + LHFN	0.671	0.654
Basic model + LHFN + SA	0.711	0.690
Basic model + LHFN + SA + TA (DSTAN)	0.752	0.728