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

Figure 4

The architecture of the proposed three pre-Training settings for zero-shot learning.