Wardy M-04 temporal fall classification
This repository contains the versioned temporal GRU ONNX artifact produced for Wardy M-04.
Artifact
| Runtime version | Files |
|---|
wardy-urfd-gru-v1 | model.onnx, metadata.json |
manifest.json contains the SHA-256 digests verified by the Wardy model manager.
Training provenance and license boundary
The model was trained for Wardy from pose sequences derived from the UR Fall Detection Dataset (URFD). URFD is provided for non-commercial research under CC BY-NC-SA 4.0. Use and redistribution of this derived model must remain compatible with that dataset license.
The source dataset and training images are not included in this repository.
Runtime boundary
Wardy feeds tracked pose sequences into this model after person detection, tracking, and pose estimation. The model output is one signal used by Wardy's state and event rules.
Limitations
This is a project baseline, not a medical device or safety guarantee. A lying pose by itself is not sufficient evidence of a fall.
