Javad Amirian
Phd Candidate at Inria
Email : javad.amirian[at]inria.fr
Address : INRIA Rennes – Bretagne Atlantique – 35042 Rennes, France
Fax : +33 2 99 84 71 71
Assistant : +33 2 99 84 22 52 (Hélène de La Ruée)
Bio
I received a master in Artificial Intelligence from Sharif University of Technology, Tehran, in 2014 and beforehand a Bachelor of Electrical Engineering from Shahid Beheshti University, Tehran, in 2012. I joined Inria Center of Rennes in January of 2018 as PhD student of prof Julien Pettre.
My PhD project is a part of CrowdBot project, an H2020 European project that aims to enable mobile robots to navigate autonomously and assist humans in crowded areas. I am developing a Pedestrian Motion Prediction system based on crowd simulation techniques.
I have been working as computer vision expert for two years in a private company in Tehran, where we designed AI tools for sport analysis and content generation. I also developed an Iranian License Plate Recognizer. Beforehand, I was leading Cyrus Robotics team, aiming to participate in Small-size league of Robocup.
You can find my cv here or visit my Linkdin profile.
My Projects
Pedestrian Motion Prediction
As a part of CrowdBot project we are developing a pedestrian motion prediction system to be used in crowded areas. To do this, we consider different interactions of the pedestrians with crowd simulation techniques to improve the prediction accuracy.
Online Camera Calibration
Sepehr is a computer vision start-up company that develop solutions in sport analysis domain. This company creates a desktop software for analyzing the games and generating video clips.
My role was developing image processing and vision algorithms and providing high-level API for UI programmers. A main processing step in this system is to estimate camera calibration based on corresponding points in soccer pitch. The vision engine has several processing modules including camera position estimation, camera pose exhaustive search for initializing the pose tracker, shot boundary detection, LM-based optimizer and etc. Besides, this product features a view synthesis module to provide stunning scenes from different views of the field and to perform camera interpolations.
Motion Planning for Simulated Autonomous Car
This was the subject of my Master thesis, supervised by Prof. Mansour Jamzad in Image processing lab of Computer Engineering Faculty of Sharif University of Technology.
I developed a motion planning technique by mixing Artificial Potential Fields and well-known RRT path planner to computer online obstacle-free paths for an autonomous car. The car was simulated and running in Webots environment. We could also deploy this motion planner on small-size soccer robots as a practical experiment.
Smart-Debug (Qt Tool)
A simple but very useful c++ tool that I’ve developed based on qCustomPlot library to ease debugging of c++ codes. It provides a super-easy api for developers to draw plots and scatters and watch the behavior of code variables just with few lines of code. I use it almost in most of my AI tools, since debugging AI algorithms is always a challenging task.
https://github.com/amiryanj/smart-debug
Small-size Soccer Robots
In 2010 we founded a small-size soccer robots team in CE department of SBU under supervision of Prof. Eslam Nazemi. A small-size soccer team consists of 6 small mobile robots playing soccer on a green carpeted field that is 9 m long by 6 m wide. Small Size robot soccer focuses on the problem of intelligent multi-agent cooperation and control in a highly dynamic environment with a hybrid centralized/distributed system.
Beside the experiencing a team-work management of seven graduate and undergraduate students, by 2012 I was designing and developing electronic boards and embedded systems of the robots. During this period we implemented a small wireless network of Zigbee nodes to enable communication between the robots and the off-the-field computer. Decision commands sent from Off-field computer were parsed and executed with ARM-based micro-controller that run a PI controller to drive the robot wheels at a desired speed and trigger the kicker module.
After 2012 I joined the software group and starting coding AI module with c++. I developed the motion planning module based on an idea in my MSc thesis. We also introduced an approach to improve the robot movements and compensating inaccuracies in robot building using a Takegi-Sugeno fuzzy method. The source codes are published in Github.
Unfortunately the project was stopped in 2016.
Publications
A complete list of my publications is available in my google scholar profile.
My Activities
I enjoy watching and playing football. My favorite club is F.C. Persepolis that has been the home of many Iranian best footballers like Ali Daei, Ali Karimi, AhmadReza Abedzadeh, etc.
