# google/youtube-8m

Starter code for working with the YouTube-8M dataset.

Repository: https://github.com/google/youtube-8m
Canonical: https://ross.abutalabs.com/products/youtube-8m
Homepage: https://research.google.com/youtube8m/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: youtube-8m
Last push: 2021-10-25T04:30:45+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3543, "days_push": 1773, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2378, forks 848 (observed 2026-08-28T04:06:42.137582+00:00)

## What it is
Starter code from Google for training and evaluating TensorFlow models on the YouTube-8M large-scale labeled video dataset. It provides an end-to-end example covering dataset reading, model training, and evaluation, including support for the Segments dataset with human-verified temporal annotations.

## Use cases
- train video classification models on the youtube-8m dataset
- learn how to read frame-level video feature files in tensorflow
- run a kaggle video understanding challenge baseline
- evaluate video models at segment-level granularity
- train models on google cloud ml platform
- export a trained video model for mediapipe inference

## When to choose
- you want a working baseline for the YouTube-8M or YouTube-8M Segments datasets
- you are participating in a YouTube-8M Kaggle challenge
- you need an end-to-end TensorFlow example for large-scale video classification

## When to avoid
- you need a general-purpose video processing toolkit rather than dataset-specific training code
- you work with modern TensorFlow 2.x or PyTorch exclusively and cannot adapt TF 1.x-era code
- you need actively maintained code - the repo has not seen recent releases

## Facets
- artifact type: dataset
- maturity: maintenance
- function: machine-learning, deep-learning, video-processing, data-science
- domain: machine-learning, computer-vision, artificial-intelligence
- platform: python, cloud
- tags: tensorflow, video-understanding, youtube-8m, starter-code, kaggle, video-classification, frame-level-features, video, gpu

## Member repositories
- google/youtube-8m (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:42.137582+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:34:54.022268+00:00, confidence not recorded.
  - readme: https://github.com/google/youtube-8m (fetched 2026-08-28T04:06:42.137582+00:00, sha 5c4e290e7dac)
  - homepage: https://research.google.com/youtube8m/ (fetched 2026-08-29T10:15:45.883421+00:00, sha dd765728abb0)
- Data as of 2026-08-30T08:39:29.467469+00:00.