3,600 movie clips and trailers curated for story-based retrieval and narrative understanding. The dataset provides video segments paired with contextual embeddings as presented in the ACCV 2020 research paper.
Use Cases
- Train cross-modal retrieval systems using the video segments and story-based text descriptions
- Evaluate narrative-aware video understanding models using the provided contextual embeddings
- Develop automated video indexing systems based on cinematic story arcs
Strengths
- 3,600 video clips sourced from a wide variety of cinematic genres
- Includes contextual embeddings for narrative-level video analysis
- Features story-based retrieval labels as defined in the ACCV 2020 publication