OroSense — Side-View Protrusion Model (CoreML)
YOLO11n-pose model specialized for side-view tongue protrusion detection. Trained on lateral oral-motor recordings from a clinical tongue ROM dataset.
Model Details
| Property |
Value |
| Architecture |
YOLO11n-pose |
| Input size |
1280 × 1280 |
| Keypoints |
4 |
| Format |
CoreML (.mlpackage) |
| Training tasks |
latRside, latLside, protRside, protLside |
| Epochs |
50 |
| Augmentation |
Heavy (rotation, scale, HSV, mosaic, mixup) |
Keypoints
| Index |
Name |
Description |
| 0 |
left_commissure |
Always occluded from side view (visibility=0) |
| 1 |
right_commissure |
Always occluded from side view (visibility=0) |
| 2 |
tongue_tip |
Annotated tongue tip position |
| 3 |
lip_tip |
Upper lip center — manually annotated reference point |
Usage (Swift / iOS)
import CoreML
import Vision
let model = try best(configuration: MLModelConfiguration())
let request = VNCoreMLRequest(model: try VNCoreMLModel(for: model.model))
Training Metrics (best epoch)
| Metric |
Value |
| Box mAP50 |
0.995 |
| Box mAP50-95 |
0.953 |
| Pose mAP50 |
0.995 |
| Pose mAP50-95 |
0.995 |
Part of OroSense
This model is one of three task-specialized models:
