Multi-label zero-shot learning with graph convolutional networks

Computer Science

• Semantic Embedding 100%
• Visual Feature 100%
• Graph Convolutional Network 100%
• Classifier 100%
• Training Data 100%
• Generalization Ability 50%
• Knowledge Transfer 50%
• Training Phase 50%
• Object Classifier 50%
• Dependent Object 50%
• Relation Graph 50%
• Attention (Machine Learning) 50%
• Word Embedding 50%
• Experimental Result 50%
• Learning Approach 50%

Keyphrases

• Global Image Feature 50%
• Local Visual Features 50%
• Attention Network 25%
• Label Semantics 25%
• Object Classifier 25%
• Label Combination 25%
• Label Co-occurrence 25%