Coordinate Frame Recovery, Per Class NMS and Null Return Modelling for Sim2Real V2X LiDAR Detection
Coordinate Frame Recovery, Per Class NMS and Null Return Modelling for Sim2Real V2X LiDAR Detection
This paper presents an approach to the V2X-Real track of the UrbanTwin Sim2Real LiDAR Challenge. The method trains a 3D object detector exclusively on synthetic data while evaluating it on unseen real roadside LiDAR from an Ouster OS1-64 unit. The final entry achieved a combined score of 0.7457, comprising a detection mAP of 0.4382 and a realism score of 0.9251. Key improvements included recovering coordinate frame mismatches, adjusting non-maximum suppression thresholds per class to avoid suppressing pedestrian groups, and modeling null returns encoded as (0,0,0) in real point clouds.
The authors report a combined score of 0.7457 on the UrbanTwin Sim2Real LiDAR Challenge by addressing three specific issues: correcting coordinate frame mismatches, adjusting non-maximum suppression thresholds per object class to handle pedestrian groups, and modeling null returns (0,0,0) present in real Ouster OS1-64 data but absent from synthetic training sets.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
combined score
0.7457
View original evidence
arXiv:2611.03940v1 Announce Type: cross Abstract: We describe our approach to the V2X-Real track of the UrbanTwin Sim2Real LiDAR Challenge, in which a 3D object detector must be trained exclusively on synthetic data and evaluated on unseen real roadside LiDAR from an Ouster OS1-64 unit, while the synthetic point clouds are simultaneously scored for distributional realism. Our final entry scores 0.7457 combined, from
Open source S2score after coordinate frame recovery
0.352–0.5147
View original evidence
a detection mAP of 0.4382 and a realism score of 0.9251. Three findings account for most of our result, and all three were diagnostic rather than architectural. First, a coordinate frame mismatch between our predictions and the evaluation ground truth was suppressing detection AP entirely; recovering it took our score from 0.3520 to 0.5147 in a single submission. Second, a single global non maximum suppression thresh
Open source S3score before coordinate frame recovery
0.352–0.5147
View original evidence
a detection mAP of 0.4382 and a realism score of 0.9251. Three findings account for most of our result, and all three were diagnostic rather than architectural. First, a coordinate frame mismatch between our predictions and the evaluation ground truth was suppressing detection AP entirely; recovering it took our score from 0.3520 to 0.5147 in a single submission. Second, a single global non maximum suppression thresh
Open source S3overall mAP after per class NMS correction
0.0751–0.4295
View original evidence
old was destroying genuine pedestrian detections, because pedestrians occur in groups and a 0.3 IoU threshold applied to a 0.7 m box suppresses distinct neighbouring people; correcting this per class raised overall mAP from 0.0751 to 0.4295. Third, approximately 31 percent of the real reference point cloud consists of null returns encoded as the literal point (0, 0, 0), a format feature entirely absent from our synth
Open source S4overall mAP before per class NMS correction
0.0751–0.4295
View original evidence
old was destroying genuine pedestrian detections, because pedestrians occur in groups and a 0.3 IoU threshold applied to a 0.7 m box suppresses distinct neighbouring people; correcting this per class raised overall mAP from 0.0751 to 0.4295. Third, approximately 31 percent of the real reference point cloud consists of null returns encoded as the literal point (0, 0, 0), a format feature entirely absent from our synth
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.03940v1 Announce Type: cross Abstract: We describe our approach to the V2X-Real track of the UrbanTwin Sim2Real LiDAR Challenge, in which a 3D object detector must be trained exclusively on synthetic data and evaluated on unseen real roadside LiDAR from an Ouster OS1-64 unit, while the synthetic point clouds are simultaneously scored for distributional realism. Our final entry scores 0.7457 combined, from
Open source S2
old was destroying genuine pedestrian detections, because pedestrians occur in groups and a 0.3 IoU threshold applied to a 0.7 m box suppresses distinct neighbouring people; correcting this per class raised overall mAP from 0.0751 to 0.4295. Third, approximately 31 percent of the real reference point cloud consists of null returns encoded as the literal point (0, 0, 0), a format feature entirely absent from our synth
Open source S4
a detection mAP of 0.4382 and a realism score of 0.9251. Three findings account for most of our result, and all three were diagnostic rather than architectural. First, a coordinate frame mismatch between our predictions and the evaluation ground truth was suppressing detection AP entirely; recovering it took our score from 0.3520 to 0.5147 in a single submission. Second, a single global non maximum suppression thresh
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org