Logo detection & recognition in sports videos (Master Thesis/ Internhips)
In this role
The goal of this project is to develop a software demonstrator to detect and recognize a given logo on sports videos. Very few systems that exist today use state of the art deep learning, and even fewer offer a real-time performance. You will have to choose relevant datasets, annotations for training and design a real-time deep learning algorithm that affords us high accuracy and real time processing. All relevant open source resources should be used in this project.
Challenges: Building robust algorithms is hard due to varying conditions under which the logo appears. Specifically, scale, color (calibration) and orientation variations. In addition, the logo can be anywhere in the video frame and may appear multiple times on the same frame. Athletes also generate partial occlusions. Finally, the real-time constraint is challenging as videos are of high resolution, processing becomes time consuming.
Project applications: Marketing, advertisement, sponsorship / Retail industry / Social media
What you will learn: You will be junior data scientist, developing your skills in machine learning (deep learning, object detection and recognition, large data handling, real-time processing architectures) and in computer vision, and will research state of the art algorithms in these fields.
Possible extensions (depends on advancement of the aforementioned work): Multiple logo recognition / Statistics and data visualization via dashboard (Tableau) / Mobile (iOS/Androïd)
Keywords: (fast) Deep learning, SIFT, cNN, region-based cNN, YOLO, transfer learning, pattern matching
In this project, the goal is to:
- Develop a logo recognition proof of concept demonstrator with real-time processing
What we offer
Join our team as intern and you will find a young, dynamic and culturally diverse working environment.
About your profile
- Required: machine learning and deep learning, image processing, computer vision
- Software engineering: Python, pytorch or tensorflow, opencv, Java
With 50+ year of history and over 900 specialists, we offer a unique spectrum of experience, skills and technical innovations.
You want to leverage your educational background, apply your infinite curiosity and your out-of-the-box thinking.
Your initiative is a chance ! We're constantly looking for talented individuals. Check our latest job opportunities !