Video dataset, MP4 (H.264)
Household tasks, egocentric
First-person video of everyday manual tasks in real homes: cooking, food preparation, dishwashing, laundry, cleaning and small repairs. Each clip has one metadata row covering the task, country, capture settings and consent.
Request access
This is an example dataset. The figures and rows on this page are synthetic and shown to illustrate the format.
- 767
- hours
- 6,000
- clips
- 350
- contributors
- 4
- countries
- 2.8 TB
- total size
Hours by task
- cooking202
- cleaning147
- food prep131
- laundry113
- dishwashing105
- repairs69.0
Hours by country
- Indonesia258
- India194
- Philippines160
- Vietnam155
Metadata columns
One row per sample, stored as Parquet. Files are MP4 (H.264).
| Column | Type | What it holds |
|---|
| clip_id | Utf8 | Stable identifier for the clip. |
| contributor_id | Utf8 | Pseudonymous contributor identifier. Split train and test on it. |
| country | Utf8 | Country where the clip was recorded. |
| task | Utf8 | cooking, food_prep, dishwashing, laundry, cleaning or repairs. |
| duration_s | Float32 | Clip length in seconds. |
| fps | Int32 | Frames per second. |
| width | Int32 | Frame width in pixels. |
| height | Int32 | Frame height in pixels. |
| has_depth | Boolean | Whether a depth track ships with the clip. |
|
Sample rows
| clip_id | consent_version | contributor_id | country | duration_s | faces_blurred | fps | has_depth | height | size_bytes | task | video_uri | width |
|---|
| clip_11c5dd74 | contributor-consent-v1 | ctr_0167 | Indonesia | 330.10 | true | 30 | true | 1080 | 330101440 | food_prep | s3://datachad-lake/datasets/household-tasks-egocentric/clips/indonesia/clip_11c5dd74.mp4 | 1920 |
| clip_ab7d2695 | contributor-consent-v1 | ctr_0236 | Vietnam | 343.39 | true | 30 | false | 1080 | 343385925 | food_prep | s3://datachad-lake/datasets/household-tasks-egocentric/clips/vietnam/clip_ab7d2695.mp4 | 1920 |
| clip_50d48c67 | contributor-consent-v1 | ctr_0071 | Indonesia | 153.48 | true | 30 |
Read the metadata
The metadata is plain Parquet. Filter it with PyArrow, DuckDB or Polars, then load only the files you picked.
import pyarrow.compute as pc
import pyarrow.dataset as ds
metadata = ds.dataset(
"s3://datachad-lake/datasets/household-tasks-egocentric/metadata/",
format="parquet",
)
rows = metadata.to_table(filter=pc.field("country") == "Indonesia")
print(rows.num_rows, rows.column_names)