Scaling Human-Object Interaction Recognition in the Video through Zero-Shot Learning

Table 7

HOIs recognition results (mAP(%)) on Charades dataset. The model sees all HOI classes in the training phase.

Method	mAP (%)
	RGB	RGB + optical flow
ActionVLAD [7]	17.6	21.0
Sigurdsson et al. [8]	18.3	22.4
CoViAR [9]	21.9	24.1
Ours	23.64	24.73