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

Table 5

The impact of the RNNs (LSTM/GRU) on the proposed system’s performance. The model is trained with 1A + 2B and tested on all data.

Method	mAP (%)
	LSTM	GRU
2Stream + WE - VLAD	19.5	19.42
3Stream + WE - VLAD	20.84	20.63
2Stream + WE + VLAD (rnn)	20.86	20.64
3Stream + WE + VLAD (rnn)	21.27	21.19