Pau Piracés
Spatial AI
Zurich, Zurich, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolPau Piracés is a Spatial AI engineer with 17 years of experience building computer vision and 3D reconstruction systems, now based in Zurich and currently leading Spatial AI work at Staer after a research-engineer stint at Meta. He has deep hands-on expertise in SfM, bundle adjustment and robust pose estimation—demonstrated by substantive OpenSfM and OpenGV contributions that added similarity averaging, ground-control constraints, Python bindings and RANSAC solvers. At Mapillary he improved geotagging precision with sub-second time handling, reflecting a knack for making geometric pipelines more accurate and production-ready. His background spans academic research (INRIA PhD) and teaching 3D vision, blending rigorous theory with pragmatic engineering. Colleagues rely on him for solving tricky reconstruction edge cases and for turning research ideas into reliable, open-source tooling.
17 years of coding experience
15 years of employment as a software developer
M.Sc. Computer Graphics Vision and Robotics, M.Sc. Computer Graphics Vision and Robotics at Grenoble INP - UGAUPC Universitat Politècnica de Catalunya
Github Skills (27)
algorithm10
algorithms10
c-language10
exif10
python10
image-processing10
routing-slip10
bash10
cmake10
bundle10
perl10
photogrammetry10
data-structure10
numpy10
routes10
Programming languages (4)
Github contributions (5)
mapillary/OpenSfM
Nov 2013 - Mar 2022
Open source Structure-from-Motion pipeline
Role in this project:
Back-end Developer
Back-end DeveloperContributions:8 releases, 13 reviews, 1504 commits in 8 years 5 months
Contributions summary:Pau's commits primarily involved modifying the core SfM pipeline of the project, OpenSfM, specifically with respect to the bundle adjustment process. The changes included setting default parameters and fixing some potential runtime errors. Furthermore, the user integrated new features, such as the utilization of a similarity averaging method and the incorporation of ground control point constraints, to improve the overall performance of the reconstruction.
structure-from-motionpythonpipelinemotionintrinsics
laurentkneip/opengv
Sep 2015 - Oct 2018
OpenGV is a collection of computer vision methods for solving geometric vision problems. It is hosted and maintained by the Mobile Perception Lab of ShanghaiTech.
Role in this project:Back-end Developer
Back-end DeveloperContributions:37 commits, 14 PRs, 14 comments in 3 years 2 months
Contributions summary:Pau significantly contributed to the project by wrapping the `p3p_kneip` method in Python, creating a Python wrapper for the core computer vision functionality. They added the necessary `pyopengv.cpp` file to expose the functionality via Boost.Python and also updated the CMake configuration to find the NumPy libraries. Furthermore, they added and implemented various absolute and relative pose solvers, including RANSAC implementations for robust pose estimation. This suggests a focus on integrating and extending the project's core algorithms with Python bindings.
structure-from-motionsolvingvisiondeep-learninghosted
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.