Aleksandr Lomakin
3D Scanning Technology Martyr
About Me
I am a computer vision specialist with expertise in 3D scanning, 3D reconstruction, 3D/2D data processing and deep learning. Despite having over 15 years of experience in C++ software development, I am still interested in exploring new technologies, learning new skills and solving complex interdisciplinary problems.
My career began during the postgraduate studies. While working in the laboratory of optical tomography and holography, I participated in the development of the primary national standard of ellipsometry units as a part of my PhD program. Additionally, I developed my first 3D interferometric phase-shift and Fourier-transform based structured light scanners, along with calibration and reconstruction software.
After having completed the PhD thesis, I joined Artec3D to further refine my skills in C++ and computer vision. I developed Artec Spider scanner and, I believe, played a significant role in the evolution of Artec Studio 3D software.
After that, I founded a start-up and designed the first-ever wireless scanner with interchangeable lenses and a user-friendly touch interface. Furthermore, a 3D processing and visualization software with a wide range of features was developed to provide users with a comprehensive solution for creating production-ready 3D models.
Being passionate about machine learning and artificial intelligence, I joined an automotive company to contribute to the development of self-driving cars. I was responsible for detecting, classifying, and segmenting objects around a car using 2D images and 3D LiDAR point clouds. Additionally, I introduced several MLOps functionalities for data collection and implemented a system for aggregating metrics of DL experiments.
Currently, I work as an R&D team lead for 3D scanning technologies. I am enthusiastic about combining traditional computer vision techniques with innovative deep learning methods.
Skills: C++ | OpenCL | CUDA | OpenCV | OpenGL | QT | Boost | Eigen | PCL | Open3D | CGAL | Ceres | g2o | TBB | Git | Conan | Docker | CMAKE | Bash | PyTorch | ROS2 | TensorRT | ONNX | Spark
- 3D DL Perception & Reconstruction
- Neural Rendering
- Generative AI
- Computational Geometry
PhD in Optical Sciences, 2011
All-Russian Research Institute for Optical and Physical Measurements
MEng in Biomedical Engineering (Diploma with Honors), 2004
Bauman Moscow State Technical University
Experience
- Over 100x accelerated 3D sensor calibration by re-implementing and optimizing algorithms in C++ and leveraging TBB multithreading.
- Implemented an optimized structured light method to double the speed of point cloud acquisition.
- Developed a semi-supervised anomaly detection method based on the VAE approach.
- Successful supervision of a working student in research for a bachelor’s degree in the field of applied physics.
- Developed from scratch the DL solution for lane detection and classification based on a modified YoloP approach.
- 5x accelerated the lane detection neural network through the use of multithreaded C++, CUDA and TensorRT.
- Implemented a real-time algorithm of a road lane network mapping on a LIDAR point cloud around a self-driving car using the g2o library, CUDA and ROS2.
- Introduced a dataset generation system in Apache Spark as part of the Dagster MLOps framework. This system allows operators to choose the time of a day, a season, and a geo-location.
- Implemented a MLOps metrics aggregation system for regular assessment of ML projects and experiments using OpenSearch, ClickHouse, Metabase and Grafana.
- Developed a new family of hand-held 3D scanners, featuring a user-friendly built-in touch screen, which is used in more than 60 countries around the world. This product won the iF Design Award in 2021.
- Designed the first ever wireless hand-held 3D scanner with interchangeable lenses and replaceable battery.
- Developed a complete self-sufficient software suite for calibration, efficient 3D data processing, and measurements.
- Developed from the ground up a brand-new 3D scanner Artec Spider resolving a variety of problems from the optics design and calibration to metrology software and an original patented 3D reconstruction method.
- Improved the reconstruction quality of the Artec Eva 3D scanner by optimizing the reconstruction algorithm and rethinking the calibration process which made it possible to scan shiny black objects.
- Contributed to the evolution of Artec Studio 3D software by developing multiple 3D data reconstruction, registration, texturing algorithms, and various 3D measurement tools.
- Implemented a certification of 3D scanners according to the VDI/VDE 2634 standard to promote the metrological reliability of the brand for industrial use.
- Developed a 3D interferometric micro-measurement scanner with a versatile research software suite for phase shifting and Fourier transform phase reconstruction, closed-loop fringe pattern demodulation and various metrology measurements.
- Developed in collaboration with EGSolutions (Italy), an intraoral structured light 3D scanner based on the Fourier transform technique with a feature-rich GUI application for reconstruction and hardware calibration.
- Developed the national primary standard for ellipsometric angle units as part of the PhD program.
- Contributed to the development of tomography research software based on Fourier interferometry using Matlab and C++, by introducing the regularized Radon transform method applied to interferometric images.
