Table 4 Comparison of action recognition... | Emerald Publishing

Table 4

Comparison of action recognition results on the NW-UCLA dataset with supervised learning approaches.

Models	Architecture	NW-UCLA
HBRNN-L [7]	RNN	78.5
SK-CNN [23]	CNN	86.1
VA-LSTM [49]	LSTM	70.7
Denoised-LSTM [4]	LSTM	80.3
SimCLR [1]	GRU	84.4
MoCo v2 [13]		78.1
BYOL [12]		30.7
SimSiam [3]		61.4
Rand-S	GRU	83.9
MS²L [21]		85.2
EMS²L		85.8
SimCLR [1]	GCN	75.7
MoCo v2 [13]		76.4
BYOL [12]		72.7
SimSiam [3]		74.8
Rand-S	GCN	84.2
EMS²L	GCN	88.0

Models	Architecture	NW-UCLA
HBRNN-L [7]	RNN	78.5
SK-CNN [23]	CNN	86.1
VA-LSTM [49]	LSTM	70.7
Denoised-LSTM [4]	LSTM	80.3
SimCLR [1]	GRU	84.4
MoCo v2 [13]		78.1
BYOL [12]		30.7
SimSiam [3]		61.4
Rand-S	GRU	83.9
MS²L [21]		85.2
EMS²L		85.8
SimCLR [1]	GCN	75.7
MoCo v2 [13]		76.4
BYOL [12]		72.7
SimSiam [3]		74.8
Rand-S	GCN	84.2
EMS²L	GCN	88.0