June 2020 - unril present

A Senior Research Scientist at Humanising Autonomy in London, UK. My Research activities lie in four domains: Object Tracking, crossing/collision warning, smartcity and human emotional understanding.

• • I proposed a crossing/collision model for  advanced driver-assistance system. 

  • I improved  a 3D kalman filter for distance/velocity estimation.

  • I proposed a method called "cascade Matching" for object tracking.

  •  I put forward a pipeline for low-level features extracted from human beings for emotional understanding.

  • I created three models for CCTV cameras in a smartcity: Zone of interest identification, object counting and traffic categorisation.

April 2019 - March 2020

A Research Scientist in Computer Vision at Shapes AI in London, UK. My Research activities lie in two domains: Visual Reasoning, Explainable AI and 3D Search:

• • I have worked on a research project whose aim is to match 2D photos (taken by mobile cameras) with CAD models. I proposed a method via transformation networks to match features of 2D images with features of CAD models.

  • Deep Visual Reasoning for Video Understanding: With two researchers at Shapes AI, we have proposed a Deep Visual Reasoning approach for Video Understanding (detecting events such as violence, commotion and road hazard) using Deep Learning, Explainable AI (via Automata and Functional Programming).

September 2017 - September 2018

A post-doctoral fellow in Computer Vision at the GREYC laboratory in France. My post-doctorate research is entitled Structure-preserving graph signal processing : Editing of 3D colored meshes. A full description of the postdoc is available here.  This postdoctoral fellowship was part of the GRAPHSIP project which was composed of 30 members. GRAPHSIP was funded by the French National Research Agency. 

In this post-doctoral project, I tackled the problem of p-Laplacian regularization of signals on directed graphs. This problem is considered for images and 3D colored meshes’ simplification/denoising.

• • I put forward three p-Laplacians dedicated to directed graph signals and presented their results (to illustrate the effect of the proposed p-laplacians on images and 3D colored meshes) at the Graph Signal Processing Workshop. This work has also been accepted in the International Symposium on Visual Computing (ISVC 2018). 

  • I proposed a new reweighted p-Laplacian regularization of graph signals on directed graphs. This work will be submitted soon.

  • I was a member of the Organizing Committee of the 12th International IAPR workshop on Graph-Based Representations in Pattern Recognition that will be held in Tours (France), from June 19 to June 21, 2019: https://gbr2019.sciencesconf.org

September 2016 - August 2017

• • A post-doctoral fellow in Pattern Recognition at the LIFAT laboratory in Tours. I focused on the classification problem in graph space in order to keep the structural information of the images/patterns. The aim of the postdoctoral project was to propose a fast classification approach in graph space.

    • I proposed a fast k-nearest neighbors approach in graph space that has a remarkable improvement of classification time and rate especially when the number of graphs in the train set is tremendous The proposed method's code is available here.

    • With a postdoctoral fellow at the LIFAT laboratory, we put forward a proof-of-concept platform, named Photo(Graph) Gallery, that allows to visually interpret graph classification results.

    • I was one of the main organizers of a Graph Edit Distance Contest whose results were presented at the International Conference on Pattern Recognition (ICPR 2016). 

  • I was part of the NeuroGeo project, which aims at creating an interactive 3D image segmentation system. I thus designed a website for the project using PHP and Javascript with an online interactive 3D image segmentation system (using Papaya viewer) that shows the segmentation results obtained by a PhD student working on image segmentation.  

  • [IMPORTANT NOTE]: the Depth First algorithm has been integrated in the networkX library, please take a look  here 

October 2012 - May 2016

A Ph.D candidate in Pattern Recognition. My PhD thesis is entitled : Anytime and Distributed Approaches for Graph Matching.

• I proposed exact, anytime, parallel and distributed Graph Edit Distance approaches. These approaches are optimized in terms of memory and running time. Most of the codes can be found here.

•I proposed a GED benchmark to evaluate the accuracy and the scalability of Graph Edit Distance methods. For more information: http://www.rfai.li.univ-tours.fr/PublicData/GDR4GED/home.html.

• With three researchers from the LITIS laboratory of Rouen, France, we proposed two binary linear programming formulations to tackle the the Graph Edit Distance problem. The codes of the proposed formulations can be found here. These formulations were the winners in the Graph Edit Distance contest in terms of precision.

February 2012 - June 2012

A master student in Artificial Intelligence. My master thesis tackled the classification and exploration of collections of web designs. 

• • I implemented different algorithms to measure the similarity between webpage layouts (including the 2-Dimensional Dynamic Warping) in order to propose a web gallery for web designers.

  • I wrote a scientific paper that was accepted later in 2014 at the Nordic Conference on Human-Computer Interaction