Mathieu U.

Participation in a solo Kaggle competition for the recognition of American Sign Language (ASL) words from pose estimation keypoints:

- Implementation of Transformers with TensorFlow

- Use of TensorFlow LITE for mobile use of the algorithm

- Feature Engineering

- Ranked in the top 20% of participants

Mission Description:

As part of a project for a prestigious French luxury house, I was responsible for the development and implementation of a 3D human pose estimation system. This technological initiative aimed to create an immersive augmented reality experience, integrating the dimensions of fashion and technology.

Key Responsibilities and Achievements:

- Design of the capture device consisting of two cameras

- Implementation of robust calibration algorithms and procedures for precise and synchronized stereoscopic use

- Implementation of a real-time Pose Estimation algorithm

- Integration into an Augmented Reality environment (Unity)

- Design and development of a native macOS application to facilitate the launch and control of the pose estimation algorithm. This user-friendly interface was created for non-technical users.

Testing and Optimization:

Carried out numerous tests and adjustments to optimize the algorithm's performance in real-world conditions, meeting the client's quality and accuracy requirements.

Academic Project: Vehicle Tracking

- Development of a Classification and Vehicle Tracking algorithm for real-time use