Heading Estimation on UAV-UGV Alignment dataset Offline (test)
0.1506Heading Error (MAE)YOLO-ANN pipeline
Evaluation results
| Method | Links | |
|---|---|---|
| YOLO-ANN pipelineHardware=RTX 3060 + Intel Core i7-12700K + 32 GB RAM, Inference Latency=31 ms / frame, Throughput (FPS)=33–35 FPS, Sensor Dependency=Monocular camera only, Tracking Modality=Heading regression per frame, Pose / Heading Source=ANN regression from bounding-box geometry (markerless), Marker Dependency=None, Dataset Scale=9,000 images total, Model Size=YOLOv5 ~7.2M Parameters (~14MB), Multiple Vehicle Detection & Tracking=Multiple UGV, Computational Footprint=Lightweight single-class YOLOv5 + small ANN (~17 MB)2025.12 | 0.1506 |