SafeSora: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset.
Josef DaiTianle ChenXuyao WangZiran YangTaiye ChenJiaming JiYaodong YangPublished in: CoRR (2024)
Keyphrases
- text generation
- human activities
- video data
- video sequences
- natural language descriptions
- video dataset
- human actions
- database
- video frames
- real time
- video streams
- video search
- static images
- text detection
- news video
- weakly labeled
- text retrieval
- video content
- video images
- human subjects
- multimedia
- text mining
- action recognition
- video surveillance
- multimedia documents
- video clips
- keywords
- word level
- personality traits
- text information
- feature set
- space time
- video shots
- video database
- video analysis
- textual descriptions
- video segments
- text documents
- lecture videos
- natural language
- image sequences
- natural scene images
- multimedia data
- combining information from multiple