Transferable Feature Representation for Visible-to-Infrared Cross-Dataset Human Action Recognition

Table 5

Average recognition precision (%) comparison of the proposed method CDFAG with state-of-the-art methods. For each method, the best accuracies of various T_train values are shown. The best accuracies of each T_train are marked in italics.

Method	T_train = 5	T_train = 10	T_train = 15	T_train = 20	T_train = 25
KEMA [15]
SSMA [16]
DA [17]
WSCDDL [18]
Dual [19]
No Adaptation
Baseline
CDFAG	50.75 ± 2.19	62.42 ± 2.59	68.08 ± 3.87	71.58 ± 2.42	75.42 ± 3.13