Songyou Peng
Research Scientist at Google DeepMind
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolSongyou Peng is a research scientist at Google DeepMind with a decade of experience building 3D/4D scene representations, pose estimation, and multi-modal spatial reasoning systems. He co-led world-scale 3D scene representation work in collaboration with Google Maps and contributes to Gemini to improve spatial reasoning for multi-modal LLMs. His academic path includes a PhD from ETH Zürich under Marc Pollefeys and Andreas Geiger, earning the ECVA PhD Award 2024, and he has a track record of mentoring PhD and master students. Songyou’s internship at Google produced OpenScene (CVPR 2023), which seeded Google’s Geo Foundational Features effort, and his open-source contributions include practical dataset and evaluation improvements to convolutional occupancy networks. Equally comfortable in research and engineering, he has repeatedly turned novel geometry ideas into scalable systems used in industry projects. Based in San Francisco, he blends deep theoretical grounding with hands-on implementation across academia and leading research labs.
10 years of coding experience
2 years of employment as a software developer
Master of Science (MSc) Computer Vision and Robotics (VIBOT), Master of Science (MSc) Computer Vision and Robotics (VIBOT) at Technical University of MunichHeriot-Watt University Edinburgh CampusMaster of Science (MSc) Computer Vision and Robotics (VIBOT), Master of Science (MSc) Computer Vision and Robotics (VIBOT) at Université de BourgogneBachelor of Engineering (B.Eng.) Automation, Bachelor of Engineering (B.Eng.) Automation at Xi'an Jiaotong UniversityDoctor of Philosophy (PhD) Computer Vision and Machine Learning, Doctor of Philosophy (PhD) Computer Vision and Machine Learning at ETH ZürichMaster of Science (MSc) Computer Vision and Robotics (VIBOT), Master of Science (MSc) Computer Vision and Robotics (VIBOT) at Universitat de Girona
Chinese, English, German
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (10)
computer-vision10
deep-learning10
python10
3d-reconstruction10
3d10
data-processing9
point-cloud9
pytorch8
scikit7
scikit-learn7
Programming languages (6)
Github contributions (5)
autonomousvision/convolutional_occupancy_networks
Jul 2020 - Jul 2022
[ECCV'20] Convolutional Occupancy Networks
Role in this project:
ML Engineer
ML EngineerContributions:17 commits, 1 PR, 12 pushes in 2 years
Contributions summary:Songyou focused on modifying and updating the `build_dataset.py` script, likely involved in generating and processing data for training. They added a feature to remove walls during mesh evaluation within `eval.py` and modified `src/common.py` to correct a bug. The commits reflect improvements to the dataset creation process and evaluation procedures, indicating a focus on model training and performance.
kitti-dataseteccvoccupancycomputer-vision3d-reconstruction
pengsongyou/msc-thesis
Apr 2017 - Jul 2017
Contributions:40 commits, 6 pushes in 2 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.