This project aims to enhance object detection and tracking algorithms by integrating scene geometry into the detection process. Utilizing mesh data of the scene and video footage from one or multiple cameras, the project seeks to develop a detection system that can leverage this additional information to improve accuracy and efficiency.
Project Overview
The integration of scene geometry into people detection algorithms offers a promising avenue to overcome some of the traditional challenges faced in crowded or complex environments. By understanding the layout and structure of the environment, the detection system can make more informed predictions, potentially reducing false positives and improving tracking consistency.
Objectives
– To develop a people detector that utilizes scene geometry for improved prediction accuracy.
– To explore and evaluate different approaches for integrating scene geometry into detection algorithms, including 3D voxel representations and 2D re-projections.
Prerequisites
Ideal candidates for this project should have:
– Strong background in computer vision and machine learning.
– Experience with 3D data processing and representation.
– Proficiency in programming, preferably in Python, and familiarity with PyTorch deep learning framework.
– Understanding of camera models and scene geometry.
Supervision
This project is co-supervised by Julien Pilet and Carlos Becker at Invision AI, in Renens, and by Martin Engilberge at CVLab, EPFL.
Contact
Please send an email to [email protected] and [email protected] if you are interested in this project.