An Investigation into Performance Factors of Two-Stream I3D Networks

Computer Science

• Backbone Network 50%
• Convolutional Network 50%
• Data Augmentation 50%
• Data Cleansing 50%
• Data Preprocessing 50%
• Feature Extraction 50%
• Network Architecture 50%
• Performance Factor 100%
• Pre-Trained Model 50%
• Recognition Accuracy 50%
• Residual Neural Network 50%
• Training Data 50%
• Video Action Recognition 50%

Keyphrases

• 3D ConvNets 100%
• Architecture Refinement 16%
• Backbone Model 16%
• Convolutional Networks 16%
• Data Augmentation 33%
• Data Cleansing 16%
• Data Preprocessing 16%
• Experiment Results 16%
• Feature Classification 16%
• Human Action Recognition 16%
• Improvement Efforts 16%
• Improvement Strategies 16%
• Key Performance Factors 16%
• Network Architecture 16%
• Network Backbones 16%
• Performance Factor 100%
• Proposer 16%
• Recognition Accuracy 16%
• ResNet50 16%
• Spatio-temporal Feature Extraction 16%
• Training Data 16%
• Two-stream 100%

Engineering

• Data Augmentation 50%
• Data Preprocessing 50%
• Feature Extraction 50%
• Improvement Strategies 50%
• Input Data 50%
• Performance Factor 100%
• Recognition Accuracy 50%
• Resnet 50%