CV
General Information
Full Name | Vardeep Singh Sanhu |
Date of Birth | 14th April 1996 |
Languages | English, German, Punjabi |
Experience
-
Jan 2023 - Present
Perception Engineer
Motor AI GmbH, Berlin
- Camera-Lidar-Radar Fusion Model - Low-Level Latent Fusion Model for Multi-view camera, LiDARs, and RADARs to solve 3D Detection and Lane Segmentation. - Implemented TensorRT plugins for unsupported operation. CUDA accelarations for parallelizable operations. Used C++, CUDA, TensorRT, PyTorch.
- Monocular 3D Detection - Developed a 3D Object Detection model on monocular images. Developed TensorRT engine and inference in C++, CUDA. - Deployed on the test vehicle achieving 3x inference speedup after quantization.
- Lidar based 3D Detection - Deployed LiDAR-only model for 3D detection. Developed TensorRT engine resulting in 24 FPS inference speed on the test vehicle.
- Face and Number Plate Detection and Blurring Developyed and deployed the model at 100FPS on test vehicle.
-
May 2022 - Nov 2022
Master Thesis Student
CARIAD SE, Mönsheim
- RADAR-based Moving Object Segmentation
- Developed a novel temporal transformer network.
- Utilized in-house Porsche RADAR Dataset along with RadarScenes dataset.
- Achieved state-of-the-art results with 12% improvement over baseline.
- Published research paper in ICRA under the guidance of Prof. Dr. Cyrill Stachniss and Uni-Bonn IPB Lab.
- Used Python libraries like PyTorch, Numpy, Pandas, and Open3D.
-
Dec 2022 - Apr 2022
Computer Vision Intern
Robidia GmbH, Cologne
- Developed an identity tracking and motion prediction suit for the camera slider.
- Implemented Kalman Filter-based motion prediction of the person-of-interest.
- Deployed the Deep Learning model on NVIDIA edge device at 60 FPS.
-
Jul 2021 - Nov 2021
Deep Learning Intern
SCREWERK GmbH, Cologne
- Developed a deep network to classify camera images based on the density of screws.
- Established WebSocket connection of frontend UI with the Deep Learning model from scratch.
- Resulted in full automation of the machine.
-
Aug 2020 -- Jun 2021
Graduate Student Assistant, HiWi
Stachniss Lab, University of Bonn
- Worked on sensor synchronization and calibration of the data collection vehicle.
- Researched classical 3D CV approaches to removal of dynamic objects from LiDAR scan.
Education
-
2022
M.Sc in Mobile Sensing and Robotics.
University of Bonn, Bonn, Germany
- Studied courses like Computer Vision, Techniques of Self-Driving Cars, Mobile Sensing Robotics, Sensor and State Estimation, Deep Learning for Visual Recognition, and Point Cloud Processing.
- Took additional elective courses like Modern C++ for Computer Vision and Vision System which provided practical PyTorch and C++ experience.
- Grade - 1.4
-
2017
B.Tech. in Electronics and Communications Engineering
Guru Nanak Dev University, India
- The course covered essential topics in Embedded Systems, Microelectronics, and mathematical topics like Linear Algebra, Statistics, and Higher-Order Calculus.
- I also took elective courses from the Computer Science Department like Computer Architecture and Fundamentals in Computer Sciences.
- Grade - 1.7
Skills
-
Development.
- C++.
- CUDA.
- OpenCV.
- LibTorch.
-
Deployment.
- TensorRT Quantization and Plugins.
- NVIDIA Orin Driveworks, DRIVE OS.
- Docker.
- CI/CD Pipelines.