Wardy M01 Person Detector
Person detection model for the Wardy patient safety monitoring system.
Model information
- Task: Person detection
- Class: person
- Input size: 640 x 640
- Training epochs: 50
- Intended tracker: ByteTrack
- Deployment target: NVIDIA Jetson with TensorRT
Files
- best.pt: Ultralytics person detector model
- training_args.yaml: training configuration
- evaluation/results.csv: training metrics by epoch
- evaluation/results.png: training curves
- evaluation/confusion_matrix.png: confusion matrix
- evaluation/val_batch0_labels.jpg: ground-truth labels
- evaluation/val_batch0_pred.jpg: model predictions
Python usage
Install Ultralytics and load best.pt with the YOLO API.
Tracking
Use best.pt with ByteTrack to maintain a track ID for each person across video frames.
Deployment
Export best.pt to ONNX, then build the TensorRT engine on the target Jetson device.
TensorRT engine files are device and TensorRT-version dependent and are not included.
Version
Initial release: v1.0.0
Limitations
This model detects people but does not identify individual identities or determine falls by itself.
- Downloads last month
- 19