Mixup Mask Adaptation: Bridging the gap between input saliency and representations via attention mechanism in feature mixup

Keyphrases

• Adaptation 100%
• Bridging the Gap 100%
• Saliency 100%
• Attention Mechanism 100%
• Boundary Issues 42%
• Input Level 42%
• Hard Boundary 42%
• Feature Map 28%
• Feature Level 28%
• Performance Improvement 14%
• Salient Region 14%
• Performance Gain 14%
• Overfitting 14%
• Mismatch Problem 14%
• Corruption 14%
• Unseen 14%
• Sharp Edge 14%
• Saliency Based 14%
• Comprehensive Experiment 14%
• Regularization Method 14%
• Deep Neural Network 14%
• Attention-based 14%
• Data Augmentation 14%
• Saliency Aware 14%
• Saliency Masks 14%
• Softmax 14%
• Robustness Enhancement 14%
• Model Generalization 14%
• Feature Augmentation 14%
• Spatial Significance 14%

Computer Science

• Attention (Machine Learning) 100%
• Feature Map 100%
• Salient Region 50%
• Performance Gain 50%
• Regularization 50%
• Performance Improvement 50%
• Data Augmentation 50%
• Inherent Complexity 50%
• Deep Neural Network 50%

Engineering

• Bridging 100%
• Input Level 100%
• Level Feature 66%
• Regularization 33%
• Performance Improvement 33%
• Input Image 33%
• Deep Layer 33%
• Deep Neural Network 33%