A Cross-View Gait Recognition Method Using Two-Way Similarity Learning

Table 1

Parameters of the TSLN module.

	Layer	Input	Output	Activation function
Temporal motion constraint difference extractor	LSTM-1	88 × 50	88 × 50	Tanh
	LSTM-2	88 × 50	88 × 50	Tanh
	Flatten		4,400
Spatio action constraint difference extractor	Residual block	See ResNet-50 parameters		ReLU
	Flatten		4,096
Spatiotemporal difference features Extractor	FC-1	8,496	1,024 (dropout)
	FC-2	1,024	128 (dropout)
	Softmax	128	2