Vision Analysis

Vision Analysis is the website and data repo for public object-detection benchmark records.

The benchmark harness lives in vision-analysis-benchmark. This repo stores reviewed submission JSON files, builds one canonical benchmark dataset, and serves that dataset through the website.

Development

cd website
npm install
npm run dev
npm run build
npm run lint

Data Flow

vision-analysis-benchmark
  -> contributor runs benchmark locally
  -> submission JSON
  -> pull request adding submissions/*.json
  -> scripts/validate_submission.py
  -> scripts/build_verified_results.py
  -> generated/verified-results.v1.json
  -> website loader/fetcher

Repo Layout

schemas/
  submission.v1.json
submissions/
  *.json
generated/
  verified-results.v1.json
scripts/
  validate_submission.py
  build_verified_results.py
website/
  src/data/metadata/
  src/lib/data/

Submission Flow

1. Run the harness from vision-analysis-benchmark.
2. Add the emitted JSON file under submissions/.
3. Open a pull request.
4. CI validates the schema and support matrix.
5. Commit the rebuilt generated/verified-results.v1.json.

Commands

Validate submissions:

python3 scripts/validate_submission.py

Rebuild canonical results:

python3 scripts/build_verified_results.py

Run the site:

cd website
npm run dev

Notes

• The website reads only generated/verified-results.v1.json.
• The website currently publishes raw benchmark metrics, not a composite score.
• Model, family, runtime, and hardware metadata still live under website/src/data/metadata/.