Teaching Yourself: A Self-Knowledge Distillation Approach to Action Recognition

Keyphrases

• Action Recognition 100%
• Self-distillation 100%
• Lightweight Network 50%
• Teacher Networks 50%
• Distillation Method 50%
• 3D ResNet 50%
• Knowledge Distillation 50%
• Neural Network 25%
• Recognition Performance 25%
• Supervised Learning 25%
• Training Process 25%
• ResNet50 25%
• Backbone Network 25%
• Effective Techniques 25%
• Complex Knowledge 25%
• HMDB51 25%
• Human Action Recognition 25%
• UCF-101 25%
• Generalization Ability 25%
• Still Images 25%
• Ablation Study 25%
• Capability Model 25%
• Augmentation Strategy 25%
• Trained Teachers 25%
• Consistency Training 25%
• Network Generalization 25%
• ResNet18 25%
• Distillation Technique 25%
• Distilling 25%
• Dark Knowledge 25%
• Overconfident Prediction 25%
• Semi-supervised Method 25%
• Student Networks 25%
• Data Diversity 25%
• Input Variability 25%

Computer Science

• Action Recognition 100%
• Knowledge Distillation 100%
• Experimental Result 16%
• Neural Network 16%
• Recognition Performance 16%
• Residual Neural Network 16%
• Supervised Learning 16%
• Training Process 16%
• Backbone Network 16%
• Semisupervised Learning 16%
• human action recognition 16%
• Data Diversity 16%
• Trained Teacher 16%