Wardy
Collection
Jetson-based on-device AI for home safety: person tracking, identity, pose, fall and hazard detection, plus the hazard-object dataset. • 6 items • Updated • 1
Wardy M-02 registered-subject identification
This repository pins the two OpenCV Zoo ONNX artifacts used by Wardy M-02 for local registered-subject identification.
Artifacts
| Component | Runtime version | Path | Upstream license |
|---|---|---|---|
| YuNet face detector | opencv-yunet-2023mar |
yunet/model.onnx |
MIT |
| SFace face recognizer | opencv-sface-2021dec |
sface/model.onnx |
Apache-2.0 |
Each component directory includes a Wardy manifest.json containing the expected SHA-256 digest.
Runtime boundary
Wardy uses these models locally with OpenCV. Registered face references, embeddings, review images, camera frames, and identity results are stored only on the user's device and are not included in this repository.
Provenance
- YuNet is redistributed from the OpenCV Zoo
face_detection_yunetdirectory under its MIT license. - SFace is redistributed from the OpenCV Zoo
face_recognition_sfacedirectory under its Apache-2.0 license. - Wardy does not claim authorship of either upstream model.
Limitations
Identification quality depends on registration images, lighting, pose, occlusion, and camera quality. This project baseline must not be treated as biometric authentication or a medical safety guarantee.