Per-detection Classification on BillboardLamac (test)
0.681Accuracyautomated pipeline for billboard detection and driver gaze duration estimation
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| automated pipeline for billboard detection and driver gaze duration estimationBounding box position (B)=true, DINOv2 full image features (Ifull)=true2026.01 | 0.681 | 0.474 | 0.629 | |
| automated pipeline for billboard detection and driver gaze duration estimationBounding box position (B)=true, DINOv2 full image features (Ifull)=true, DINOv2 cropped image features (Icrop)=true2026.01 | 0.673 | 0.483 | 0.625 | |
| automated pipeline for billboard detection and driver gaze duration estimationBounding box position (B)=true, DINOv2 cropped image features (Icrop)=true2026.01 | 0.638 | 0.364 | 0.555 | |
| automated pipeline for billboard detection and driver gaze duration estimationDINOv2 cropped image features (Icrop)=true2026.01 | 0.632 | 0.259 | 0.49 | |
| automated pipeline for billboard detection and driver gaze duration estimationDINOv2 full image features (Ifull)=true, DINOv2 cropped image features (Icrop)=true2026.01 | 0.632 | 0.277 | 0.502 | |
| automated pipeline for billboard detection and driver gaze duration estimationDINOv2 full image features (Ifull)=true2026.01 | 0.63 | 0.263 | 0.49 | |
| automated pipeline for billboard detection and driver gaze duration estimationBounding box position (B)=true2026.01 | 0.622 | 0.391 | 0.561 |