A Coarse-to-Fine Dual Attention Network for Blind Face Completion

Keyphrases

• Reconstruction Accuracy 100%
• Coarse-to-fine 100%
• Dual Attention Network 100%
• Face Completion 100%
• Occluded Region 66%
• Occluded Faces 66%
• Occlusion 33%
• Facial Landmarks 33%
• Face Features 33%
• Sparsity 33%
• Feature Map 33%
• Patch-wise 33%
• Pixel-wise 33%
• Information Exchange 33%
• Conditional Generative Adversarial Network (cGAN) 33%
• Semantic Content 33%
• Cross Attention Module 33%
• Feature Extractor 33%
• Closing the Gap 33%
• True Positive 33%
• Adversarial Loss 33%
• Identification Rate 33%
• Face Recognition 33%
• Positive Identification 33%
• Global Self-attention 33%
• Landmark Prediction 33%
• Quality Recognition 33%
• Coarse-to-fine Network 33%
• Occluded Area 33%
• Rough Estimation 33%
• Parallel Attention Mechanism 33%

Computer Science

• Attention (Machine Learning) 100%
• Sparsity 50%
• Generative Adversarial Networks 50%
• Trained Network 50%
• True Positive 50%
• Face Recognition 50%
• Missing Information 50%
• Structural Similarity 50%

Neuroscience

• Face 100%
• Face Perception 14%