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

Table 1

Top-1 accuracy of different models on the AWA2 and aPY datasets in large-scale setting using the ImageNet dataset.

	Method	AWA2 (%)	aPY (%)
OL	Con	44.5	26.4
	DeViSE	59.7	37.0
		61.0	35.0
	SE-	69.2	—
		70.7	—
		77.3	91.
	AGCNZ	79.0	91.2
	ADGPZ	80.3	90.4
ML		81.7	91.5
		78.3	91.4
		82.1	90.6