Table 5 Comparison of action recognition...

Table 5

Comparison of action recognition results on the PKUMMD dataset with supervised learning approaches.

Models	Architecture	Part I	Part II
ST-GCN [47]	GCN	84.0	48.2
VA-LSTM [49]	LSTM	84.1	50.0
SimCLR [1]	GRU	87.0	44.8
MoCo v2 [13]		86.6	43.7
BYOL [12]		68.1	38.4
SimSiam [3]		85.6	44.1
Rand-S	GRU	81.5	44.1
MS²L [21]		83.4	42.4
EMS²L		86.4	46.6
SimCLR [1]	GCN	85.2	38.2
MoCo v2 [13]		82.3	34.9
BYOL [12]		76.6	38.9
SimSiam [3]		86.4	41.4
Rand-S	GCN	86.8	42.4
EMS²L	GCN	86.8	45.2

Models	Architecture	Part I	Part II
ST-GCN [47]	GCN	84.0	48.2
VA-LSTM [49]	LSTM	84.1	50.0
SimCLR [1]	GRU	87.0	44.8
MoCo v2 [13]		86.6	43.7
BYOL [12]		68.1	38.4
SimSiam [3]		85.6	44.1
Rand-S	GRU	81.5	44.1
MS²L [21]		83.4	42.4
EMS²L		86.4	46.6
SimCLR [1]	GCN	85.2	38.2
MoCo v2 [13]		82.3	34.9
BYOL [12]		76.6	38.9
SimSiam [3]		86.4	41.4
Rand-S	GCN	86.8	42.4
EMS²L	GCN	86.8	45.2