Bringing Image Structure to Video via Frame-Clip Consistency of Object Tokens

Keyphrases

• Image Structure 100%
• Video Understanding 100%
• Object Token 100%
• Performance Improvement 50%
• No-contact 50%
• Recognition Model 50%
• Domain of Interest 50%
• Small Sets 50%
• Instantiation 50%
• Learning Framework 50%
• Action Recognition 50%
• Scene Structure 50%
• Temporal Action Localization 50%
• Still Images 50%
• Annotated Images 50%
• Point of No Return 50%
• Transformer Model 50%
• Object Graphs 50%
• Multiple Videos 50%
• Consistency Loss 50%
• Individual Representation 50%
• Video Model 50%
• Scene Representation 50%

Computer Science

• Structured Information 100%
• Video Understanding 100%
• Annotation 50%
• Learning Framework 50%
• Action Recognition 50%
• Performance Improvement 50%
• Scene Structure 50%
• Instantiation 50%