yolozu 1.0.0

Contract-first evaluation + tooling harness for detection/pose/segmentation.

YOLOZU (萬)

日本語: Readme_jp.md

Interface-contract-first evaluation harness for detection / segmentation / pose.

Run inference in any backend, export a stable predictions.json interface contract, and evaluate apples-to-apples with the same validators and metrics.

Quickstart (run this first)

bash scripts/smoke.sh

Output artifact: reports/smoke_coco_eval_dry_run.json.

Docs index (start here): docs/README.md.

Start here (choose 1 of 4 entry points)

• A: Evaluate from precomputed predictions (no inference deps) — predictions.json → validate → eval.
• B: Train → Export → Eval (RT-DETR scaffold) — run artifacts → ONNX → parity/eval.
• C: Interface contracts (predictions / adapter / TTT protocol) — schemas + adapter boundary + safe adaptation protocol.
• D: Bench/Parity (TensorRT / latency benchmark) — parity checks + pinned-protocol benchmarks.

All four entry points are documented (with copy-paste commands) in docs/README.md.

CLI note:

• yolozu ... is the pip/package CLI.
• python3 tools/yolozu.py ... is the repo wrapper CLI.
• For equivalent commands, swap only the executable (yolozu ↔ python3 tools/yolozu.py).

Key points

• Bring-your-own inference → stable predictions.json interface contract.
• Validators catch schema drift early.
• Protocol-pinned export_settings makes comparisons reproducible.
• Parity/bench quantify backend drift and performance.
• Tooling stays CPU-friendly by default (GPU optional).
• Apache-2.0-only ops policy is enforced in repo tooling.

Why YOLOZU?

• Run inference in any environment you prefer (PyTorch / ONNXRuntime / TensorRT / C++ / etc.) and save only the results to the common predictions.json interface contract.
• YOLOZU validates and scores that JSON with the same evaluator, so you can compare backend differences under identical conditions and reproduce results more easily.
• export_settings records preprocessing/protocol settings, making comparisons reproducible over time.
• Details: docs/yolozu_spec.md.

Install (pip users)

python3 -m pip install yolozu
yolozu --help
yolozu doctor --output -

Optional extras and CPU demos: docs/install.md.

Source checkout (repo users)

python3 -m pip install -r requirements-test.txt
python3 -m pip install -e .
python3 tools/yolozu.py --help
python3 -m unittest -q

Manual (PDF)

Printable manual source: manual/.

Support / legal

• Contact: [e-mail]
• © 2026 ToppyMicroServices OÜ Full support/legal: docs/support.md.

License

Code in this repository is licensed under the Apache License, Version 2.0. See LICENSE.