Attention-Based Graph Convolutional Network for Zero-Shot Learning with Pre-Training

Table 2

Top-1 accuracy of different models on the AWA2 test set.

Method	Accuracy (%)
	Original loss		Modified loss
	Classifier	All	Classifier	All
GCN	38.7	34.8	40.2	35.7
DGP	43.9	36.5	52.2	38.0
AGCNZ	49.7	36.0	50.7	37.0
ADGPZ	50.2	38.4	55.6	39.6