HERO: Hierarchical Encoder for Video+Language Omni-representation Pre-training.
Linjie LiYen-Chun ChenYu ChengZhe GanLicheng YuJingjing LiuPublished in: EMNLP (1) (2020)
Keyphrases
- representation language
- hierarchical representation
- video sequences
- video data
- video streams
- video content
- training set
- digital video
- language learning
- multimedia
- programming language
- image representation
- mpeg standard
- real time
- omni directional
- video analysis
- video surveillance
- video frames
- training process
- key frames
- multimedia data
- compressed video
- temporal correlation
- bit rate
- natural language