MovieNet: a holistic dataset for movie understanding
Source: https://movienet.github.io/
Structured fields
{
"checked_at": "2026-10-17T00:00:00.000Z",
"usefulness": "Largest public annotation set with shot scale (5 classes) and shot movement (4 classes) over 1,100 films plus scene boundaries and synopsis alignment; a basis for measuring shot-scale distribution automatically.",
"source_type": "dataset",
"availability": "available",
"source_owner": "MMLab CUHK / OpenMMLab (Huang, Xiong, Rao, Wang, Lin)",
"next_check_at": "2026-11-16T00:00:00.000Z",
"known_limitations": [
"Not updated since 2020",
"Movie video files unavailable"
],
"upstream_updated_at": "2020-10-17T00:00:00.000Z"
}Evidence claims
- [sourced_fact] Shot cinematic style: shot scale (5 categories from extreme close-up to long shot) and shot movement (4 categories such as static and zoom in). · locator: annotations section source
- [sourced_fact] 1,100 movies, 60K trailers, 3.9M photos, 42K scene boundaries, 1.1M character instances; ECCV 2020 spotlight. · locator: contents and size section source
- [sourced_fact] Downloads via OpenDataLab require accepting a User Service Agreement; movie files not released pending CUHK legal approval; site last updated July 2020. · locator: download section source
- [upstream_claim] Research review verdict at export: reviewed (as of 2026-10-17). Recorded by the research pipeline, not a hub review outcome; evidence.review stays unreviewed until an independent reviewer records one.
Rights
| component | license | hosting | redistribution | modification | training | Evidence and notes |
|---|---|---|---|---|---|---|
| dataset | unknown | restricted | restricted | unknown | unknown | Access gated by OpenDataLab agreement (not read); no open license named; underlying films are third-party copyrighted works. |