Attentive 3D-Ghost Module for Dynamic Hand Gesture Recognition with Positive Knowledge Transfer

Table 2

8-fold cross-subject average accuracy (%) of different hand gesture methods on the VIVA dataset.

	Testing modality
Method	RGB	Depth
HOG + HOG2 [28]	52.3	58.6
CNN:LRN [13]	57.0	65.0
C3D [33]	71.26	68.32
I3D [6]	77.84	74.46
MTUT [5]	80.03	78.97
Ours	81.50	80.15

In the experiment results, the best recognition accuracy is in bold.