Datasets
Browse by modality, format and tag. Every dataset page shows its size, coverage and metadata columns. Entries marked Example use synthetic figures to illustrate the format.
Loading the catalogue.
Browse by modality, format and tag. Every dataset page shows its size, coverage and metadata columns. Entries marked Example use synthetic figures to illustrate the format.
Loading the catalogue.
4 of 4 datasets
First-person street video recorded from the helmets of motorbike ride-hailing and delivery drivers.
Video / 429 hours / 5,000 clips / 2.5 TB
Unscripted speech in Southeast and South Asian languages, recorded by native speakers in everyday settings.
Audio / 262 hours / 40,000 utterances / 29.9 GB
Questions and answers written from scratch by native speakers, in languages the web under-represents.
Text / 19.1 million tokens / 30,000 documents / 95.7 MB
Head-mounted video of people cooking, cleaning and repairing things in their own homes.
Video / 767 hours / 6,000 clips / 2.8 TB