PhD thesis (2008-2012): Vision Algorithms for Rain and Trafﬁc Lights in Driver Assistance Systems [pdf, 80MB]
The research scope of my thesis is the detection of visual cues to improve scene understanding for intelligent vehicles. Part of my work is focusing on object detection (traffic lights) as well as detecting and improving visibility in degraded conditions. I mainly work on two projects including traffic light recognition and rain detection. From July to end October 2011 I also worked in the robotics institute of CMU (USA) with professor Srinivasa Narasimhan on the simulation of a new weather-free lighting device.
(thesis advisor: Fawzi Nashashibi, Robotics Centre MinesParisTech & INRIA Imara Team)
Below is a non-exhaustive listing of the researches I've been involved in:
Depth image analysis for Scene segmentation and Gestures recognition in context the of wheel-throwing pottery 
Post-doctoral position at Robitics Centre, Mines ParisTech. Analysis and segmentation of depth images to capture pottery gestures. Part of the european project i-Treasures to save the intangible cultural Heritage.
Using a collocated high speed camera and a projector, we improve the visibility of night scenes through selectively illumination of a scene, by avoiding direct lighting of falling particles.
Project conducted at Robotics Institute of CMU, with professor Srinivas Narasimhan and PhD Robert Tamburo.
> Project page (external page CMU)
When driving in daytime the effect of rain is seen both as streaks in the air and drops on the windshield. Usual vision algorithms are prone to failures due to visual distortion of the scene. We developed several new methods to tackle the detection of unfocused raindrops detection using low processing, or frequency analysis.
Real time recognition of traffic light in dense dynamics environment using spot light detection and templates matching.
> Our public benchmarks