Research map / 3D vision / Depth and stereo
Robust Automatic Monocular Vehicle Speed Estimation for Traffic Surveillance
Jerome Revaud, Martin Humenberger
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use masked IoU instead of box IoU in the reprojection error (ablation) | 3D-reproj · BrnoCompSpeed | abs. error average (km/h): 5.70 -> 2.84 | Improved |
| add RANSAC filtering of predicted Jacobians (ablation) | Learned method · BrnoCompSpeed | abs. error average (km/h): 9.21 -> 2.15 | Improved |
| add RANSAC filtering of predicted Jacobians (ablation) | Learned method · CCTV dataset | abs. error average (km/h): 20.1 -> 12.8 | Improved |
| add context from other detections via transformer attention (ablation) | Learned+RANSAC · BrnoCompSpeed | abs. error average (km/h): 3.01 -> 2.15 | Improved |
| increase number of 3D car models in catalog (ablation) | 3D-reproj · synthetic benchmark | speed error | No clear effect |
| add Gaussian noise to bounding boxes up to 10% of box size (ablation) | 3D-reproj without mask · synthetic benchmark | speed error | Hurt |
| add Gaussian noise to bounding boxes (ablation) | Learned method · synthetic benchmark | speed error | No clear effect |
| optimize speed error directly instead of mu and Jacobian l2 loss (ablation) | Learned method · synthetic training | speed error | Hurt |
| replace FullACC++ with learned synthetic-trained calibration | speed estimation · CCTV dataset | abs. error median (km/h): 22.4 -> 5.82 | Improved |
| use mask IoU reprojection instead of Learned+RANSAC | calibration · CCTV dataset | abs. error median (km/h): 5.82 -> 5.45 | No clear effect |
Closest papers by meaning
- CityFlow: A City-Scale Benchmark for Multi-Target Multi-Camera Vehicle Tracking and Re-Identification (CVPR 2019 · similarity 0.85)
- On-Road Object Importance Estimation: A New Dataset and A Model with Multi-Fold Top-Down Guidance (NeurIPS 2024 · similarity 0.84)
- TARS: Traffic-Aware Radar Scene Flow Estimation (ICCV 2025 · similarity 0.83)
- Robust Frame-to-Frame Camera Rotation Estimation in Crowded Scenes (ICCV 2023 · similarity 0.83)
- STREETS: A Novel Camera Network Dataset for Traffic Flow (NeurIPS 2019 · similarity 0.83)
- Ev-3DOD: Pushing the Temporal Boundaries of 3D Object Detection with Event Cameras (CVPR 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
