Center for Environmental Research & Technology

CE-CERT Team Wins Gold at CVPR DriveX Challenge

June 20, 2025

A CE-CERT research team has earned the Gold Award at the CVPR 2025 DriveX Workshop, an international competition focused on developing advanced

models for connected and autonomous vehicles. The workshop was held as part of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)—the top-ranked conference in the field of computer vision, according to Google Scholar.

The project was led by Chuheng Wei, a Ph.D. candidate in Electrical and Computer Engineering at UC Riverside, under the guidance of his co-advisors, Professors Guoyuan Wu and Matthew J. Barth. Together, they led Team HetoFuse to a first-place finish in the TUMTraf V2X Grand Challenge, which evaluated how well teams could detect and track objects using shared data from multiple connected vehicles. The challenge used a complex dataset featuring real-world driving conditions across both daytime and nighttime scenarios. Out of more than 80 submissions from 10 teams, HetoFuse achieved a top accuracy score of 43.38 3D mAP, ranking first among all teams except the challenge baseline.

The team’s method, called HeCoFuse, combines camera and LiDAR sensor data from multiple vehicles to build a more complete and accurate picture of the environment. This cooperative perception approach improves object detection performance in challenging scenarios by allowing vehicles to share what they “see” with one another. The method was evaluated on key metrics, including detection accuracy, position and orientation error, precision, and recall, and consistently outperformed its peers.

As part of the award, the team has been invited to present their work at the DriveX Workshop and will submit a technical report detailing their approach. The project reflects CE-CERT’s ongoing contributions to advancing real-world solutions in connected vehicle research.