上一条: GAIN: Gradient Augmented Inpainting Network for Irregular Holes
下一条: Zero-Shot Learning via Category-Specific Visual-Semantic Mapping and Label Refinement