This report presents a complete technical solution for the UrbanTwin V2X-Real Sim2Real LiDAR Challenge, covering both synthetic point cloud generation and 3D object detection. For data generation, we adopt per-axis quantile calibration. Public V2X-Real training data are used to construct the real reference distribution, and independent one-dimensional quantile mappings are estimated for the x, y, and z axes of the base synthetic point clouds. The calibrated point clouds are then resampled using a fixed random seed and clipped to the valid spatial range, producing 50 frames with 2,000 float32 XYZ points per frame. For detection, we use a two-stage Voxel R-CNN (VoxelRCNN) detector trained with RangeLDM-enhanced synthetic point-cloud data, point clouds downsampled to 50k points per frame, and reconstructed pedestrian data. Training is performed on four NVIDIA GeForce RTX 4090 GPUs with a global batch size of 16 for 40 epochs. Final testing uses direct single-model inference with the original VoxelRCNN model. On the official evaluation server, the normalized Realism score is 0.7905 and the normalized Detection score is 0.5583, with [email protected] of 0.3908, BEV [email protected] of 0.4219, and [email protected] of 0.1047. According to the official weighting rule, the final combined score is 0.6 x 0.5583 + 0.4 x 0.7905 = 0.6512. This report documents the complete implementation details of data construction, model training, inference, submission formatting, and official evaluation.
Figures & tables
Figure 1: Overall framework for synthetic point cloud generation and 3D object detection. The Realism branch applies per-axis quantile calibration, resampling, and spatial clipping to the base synthetic point clouds. The Detection branch trains VoxelRCNN with RangeLDM-enhanced synthetic point-cloud data, point clouds downsampled to 50k points per frame, and reconstructed pedestrian data, followed by direct single-model inference on the test set.
Configuration
Setting
Detector
VoxelRCNN (GBlobs fork)
Training hardware
4 × NVIDIA GeForce RTX 4090
Global batch size
16 (4 / GPU)
Training epochs
40
DataLoader workers
8
Random seed
Fixed
Table 1: VoxelRCNN training and inference configuration.
Metric
Official Result
Normalized Realism
0.7905
CD
6.90
MMD
1.25×10−2
EMD
2.26
FPD
9.34
Table 2: Official Realism evaluation results for the synthetic point cloud submission.
Lidar-based 3D object detection locates and classifies objects in point clouds, supporting perception for autonomous vehicles, robots, and other mobile systems.
We present our solution to the LUMPI track of the UCF UrbanTwin Sim2Real LiDAR Challenge at the 6th DriveX Workshop, ECCV 2026. The detector must be trained only on synthetic data and is evaluated on 50 held-out real LiDAR frames; a separate 50-frame synthetic submission is evaluated for point-cloud realism. Our method addresses the Sim2Real gap at three levels. First, we align synthetic scans to the 50k-point test density and build a 30k-record training pool using UT-LUMPI geometry, RangeLDM-based sampling diversification, rare-class copy-paste, and pedestrian-oriented augmentation. Second, complementary DSVT detectors and Car/Bus PointPillars specialists are trained under the same synthetic-only constraint. Third, predictions are integrated by class-aware routing, asymmetric agreement fusion, constrained residual-recall supplementation, class-coverage auditing, and selective box-size calibration. The realism branch is optimized independently with radial-density matching, weak affine calibration, and calibrated set mixing. The final submission obtains a Combined Score of 0.4692, a Detection Score of 0.1797, a Realism Score of 0.9035, and 3D [email protected] of 0.1258.
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 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 threshold 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 synthetic clouds, which pinned Earth Mover's Distance near its floor until we modelled it. Negative results are reported at equal length, and Section 9 gives a full account of data provenance and compliance. That section also accounts for every mechanism contributing to the submitted score.
This report summarizes the work submitted by team arnav007 (submission ID 1054222) for the UrbanTwin Sim2Real LiDAR Challenge (V2X-Real track) at the IEEE BigData Cup 2026. The primary objective is to train 3D detectors for vehicles, pedestrians, and trucks using only synthetic roadside LiDAR data, and to generate synthetic scans that are realistic enough to be indistinguishable from real ones. A CARLA based pipeline is developed and calibrated to the fixed sensor site, accurately modeling the range image, including no-return beams, based on unlabelled real development scans. A combined inference detector, using multiple 3D detectors (VoxelNeXt, PV-RCNN, and SECOND) trained exclusively on synthetic data, is also proposed. Each scan is aligned to the synthetic ground plane and processed through the detectors at multiple ground heights. The resulting bounding boxes are fused using bird's-eye view (BEV) clustering, followed by class-specific post-processing to refine heading, box size, and filter low-confidence detections. The final submission achieved a combined score of 0.8947. The ablation study shows that per-scan ground leveling more than doubled the real-scan mAP, while fusing boxes across different heights and models further improved performance.
Arnav Kumar, Brijesh Soni
Department of Computer Science and Engineering, The Ohio State University, Columbus, OH, USA