Computer Vision Researcher
About the role
As a Computer Vision Researcher, you will be developing algorithms to multiple problems related to the 2D and 3D scene reconstruction of the football match. You will be solving the task of efficient and robust camera tracking using a combination of machine learning and classical optimization methods in a challenging setup based on a single view broadcast video. You will also focus on extending our current solution to the camera tracking problem in order to provide uncertainty estimation.
Moreover, you will develop and extend methods for robust and automatic image distortion correction and panoramic pitch reconstruction.
We are open to ideas and give freedom in taken approach and experimentation and expect that you will be able to work on the given tasks and deliver results with little supervision.
Full time position. B2B contract is preferred. Work from the office in the strict center of Warsaw.
Responsibilities/Tasks:
- Developing robust and efficient methods for automatic pitch rectification tasks
- Creating novel approaches to single-view camera tracking problem and improving efficiency of our current solution.
- Working on 3D reconstruction of the football match from single and multiple-viewports
- Adapting and extending existing algorithms for panoramic image stitching based on single-view broadcast video
- Adapting algorithms to the aforementioned problems to work in real-time and online scenarios.
Requirements
- At least M.Sc in mathematics or related field (CS, physics)
- Sound knowledge of the theory of single-view and n-view geometry and camera calibration
- Sound knowledge of non-linear optimization methods
- Good knowledge of statistical theory / probabilistic machine learning (e.g. state space models like Kalman FIlter)
- Good knowledge of classical computer vision algorithms
- Experience working with image/video data
- Basic knowledge of Python
- Good command of English
Preferred
- Publications in the area of computer vision or related fields
- Knowledge of modern deep learning neural network architectures
- Experience working with deep learning algorithms and frameworks (PyTorch, TF)
- Proficiency in Python
- Experience working with Python scientific libraries (numpy, scipy) and optimization methods (numba)
Tech stack:
- Python (optionally C)
- Pytorch
- AWS
- Docker
Benefits
- Dynamic start-up environment with interesting research challenges & bleeding edge technology and models (this time for real)
- Collaboration in a group with multiple PhDs and PhD students
- Opportunity to build real-world deployable machine and deep learning-based products changing the world of football
- International projects & collaborations with top football clubs & federations
- Ability to directly influence development direction of a product
- Team Kaggle competitions (we have 3 masters) & hackathons
- Office in the strict center of Warsaw