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

Figure 8

Comparison of classification accuracy of two loss functions. Among them, (a)–8(d) are the results on the aPY dataset, and (e)–(h) are the results on the AWA2 dataset.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)