Collaborative Learning for Semi-Supervised LiDAR Semantic Segmentation
B. Yang, A. Condurache. "Collaborative Learning for Semi-Supervised LiDAR Semantic Segmentation." ICML 2026.
CV
🎓 Education
- Ph.D. in Computer Vision and Autonomous Driving — Bosch Research for AI, in cooperation with University of Lübeck (Dec 2023 – Present)
- M.S. in Robotics, Systems and Control — ETH Zurich (2021-2023), GPA: 5.67/6.0
- B.Sc. in Mechanical Engineering — RWTH Aachen University (2016-2020), Dean’s List (Top 5%)
💼 Industry Experience
- Since December 2023: Industrial Ph.D. Research Scientist at Bosch Research
- Developing 3D scene understanding methods for autonomous driving, with first-author or core contributor publications at CVPR (2024, 2026), ICML 2026, AAAI 2026, WACV 2026; one additional paper under review.
- Benchmarked radar perception methods, including object detection, occupancy prediction, and self-supervised learning, on proprietary Bosch datasets.
- Co-organized and contributed to workshops on 3D foundation models for autonomous driving perception.
- Winter 2021: Research Engineer Intern at Volkswagen AG
- Developed an AR-based data augmentation pipeline for semantic segmentation, automating obstacle catalogue extraction and camera-projection-based insertion into driving scenes across diverse ground textures;
- Boosted semantic segmentation performance by +5% mIoU through photorealistic synthetic data augmentation, reducing manual annotation overhead.
- Summer 2021: Software Engineering Intern at BMW AG
- Analyzed large-scale driving data via AWS and PySpark to characterize traffic scenario distributions across China and Germany
- Developed Monte Carlo-based risk models for autonomous driving safety-in-use analysis.
- 2019-2020: Research Intern and Bachelor Thesis Student at Schaeffler AG
- Researched optimization methods for heat pump systems in electric mobility applications. Personal Patent: (Patent No. DE102022100491A1) Geometry-variable ejector.
🔬 Research Experience at University
- Summer 2023: Master Thesis Student at ETH Zurich, Autonomous Systems Lab
- Proposed a Swin-Transformer-based method for high-resolution LiDAR point cloud reconstruction from sparse low-resolution input, published at CVPR 2024, GPA: 6.0/6.0
- Mentored one undergraduate researcher, supervising continuation of ongoing LiDAR perception research
- Winter 2022: Research Project Assistant at ETH Zurich, AIT Lab for Intelligent Interactive Systems
- Designed a camera localization method to enhance 3D gaze estimation, improving spatial accuracy of gaze-to-scene mapping through refined camera pose optimization.
🛠️ Skills
- Programming: Python, C++, CUDA, Bash
- Deep Learning: PyTorch, PyTorch Lightning, TensorFlow, Hugging Face, Weights & Biases
- 3D/CV Libraries: Open3D, PCL, MinkowskiEngine, MMDetection3D, OpenCV
- Infrastructure: Linux, Docker, ROS, GPU clusters, Azure, AWS, OpenShift
- LLM: OpenAI, Claude automations, Antropic API
🏅 Honors & Awards
- Outstanding Reviewer Award — CVPR 2025
- Silver Reviewer Award — ICML 2026
- Schaeffler Top Student Award — Schaeffler AG (2020)
- Dean’s List Award every semester — RWTH Aachen University (2016–2020, top 5%)
- Scholarship recipient — Bildungsfonds, Carl-Arthur Pastor Stiftung, Hans Hermann Voss-Stiftung (2016–2018)
📄 Publications
Towards Foundation Models for 3D Scene Understanding: Instance-Aware Self-Supervised Learning for Point Clouds
B. Yang, M. Abdelsamad, et al. (2026). "Towards Foundation Models for 3D Scene Understanding: Instance-Aware Self-Supervised Learning for Point Clouds." CVPR 2026.
DOS: Distilling Observable Softmaps of Zipfian Prototypes for Self-Supervised Point Representation
M. Abdelsamad, M. Ulrich, B. Yang, et al. (2026). "DOS: Distilling Observable Softmaps of Zipfian Prototypes for Self-Supervised Point Representation." AAAI 2026.
FLARES: Fast and Accurate LiDAR Multi-Range Semantic Segmentation
B. Yang, A. Condurache. (2025). "FLARES: Fast and Accurate LiDAR Multi-Range Semantic Segmentation." WACV 2026.
TULIP: Transformer for Upsampling of LiDAR Point Clouds
B. Yang, P. Pfreundschuh, et al. (2024). "TULIP: Transformer for Upsampling of LiDAR Point Clouds." CVPR 2024.
🤝 Service and Leadership
- Teaching Assistant — Perception for Autonomous Vehicles, University of Lübeck (since 2024)
- Reviewer — CVPR, NeurIPS, ICML, ICLR, ICCV, ECCV (2024–2026)