Amir Rasouli, M.A.Sc
PhD Candidate at YorkU

     I am currently a PhD candidate at Lassonde School of Engineering at York University and a member of prof. John K. Tsotsos lab and Center for Vision Research.
     My main topic of research is in automation and robotics in particular I am interested in tasks involved in autonomous driving including traffic scene understanding, road users' behavior analysis and control.
     My other interests involve active visual search, recognition and assisstive robotics
.


Home


Research


CV


Code


Publications


Contact

Amir Rasouli, M.A.Sc
PhD Candidate at YorkU


Home


Research


CV


Code


Publications


Contact

Education:
  • B. Eng. (Hon.) Computer Systems Engineering RMIT University, 2010
  • B.A. Business Management RMIT University , 2010
  • M.A.Sc. Computer Engineering, York University, 2014
  • PhD Candidate in Computer Science at York University, In progress

Technical Skills:
  • Computer Vision: recognition, tracking, stereoscopic, localization and visual attention
  • Programming: C++, Matlab, Java, VHDL, Assembly
  • Hardware: Logic level design, VLSI, FPGA
Teaching Experience:
  • EECS 3215: Embedded Systems      2013, 14
  • EECS 1020: Introduction to COSC I (Java Programming)      2013-15
  • EECS 1011: Computational Thinking     2015,16
  • EECS 2001: Intro. to the Theory of Computation     2015-17
  • EECS 1560: Computing for Math and Statistics     2015  
Awards:
  • Auto CRC and VPAC R&D award for undergraduate thesis
  • Top Ranking Student at RMIT in years 2008 and 09
  • Nominee for Best engineering student on year 2010
  • Performance excellence in subjects Macroeconomics and Buyer Behavior
Membership:
  • Center for Vision Research (CVR)
  • NSERC Canadian Field Robotic Network (NCFRN)
  • Golden Key International Society

Amir Rasouli, M.A.Sc
PhD Candidate at YorkU


Research


CV


Code


Publications


Contact

Current Research

Pedestrians' behavioral analysis at the point of crossing









     The objective of this project is to analyze the behavior of pedestrians in various traffic scenarios. In particular we are interested in estimating the pedestrians intention of crossing and predicting whether they cross depending on the context such as street structure or driver's reaction. paper










     

As part of this project a large dataset of pedestrians in various traffic scenes is collected. 



Previous Works

Color Space Selection in  Search and Recognition













     We evaluated 20 color spaces to measure their suitability for detection and recognition applications. The evaluation was done on a large dataset of synthetic and real images.
















     We evaluated the best performing color spaces as a means of top-down control in active visual search and found that the C1C2C3 color space is the most robust color space for the purpose of visual search.


Autonomous Robotic Visual Search in Unknown 3D environments
     
     




     





       




      We incorporated visual saliency with visual object search in 3D environments. Using a practical robot, we showed that significant improvements can be achieved using visual salient cues to guide the search. paper











     





     An extension of the work was developed and tested in a simulated environment using Gazebo with a large number of 3D models. Here, individual saliency models were examined in a large cluttered office environment. 


Combined  Spatiotemporal and Boundary Information for Visual Tracking 


     










     

     In  this work we combined two methods of visual tracking, namely contour and region based trackers. The active contours are used to define the region of interest around the tracked object. This approach reduces the dependency of the region based tracker to the background information, and as a result, more accurate performance is achieved.



FlexRay Protocol for Automotive Industry
     
     







     
     This project was sponsored by Holden Australia to examine the performance of the FlexRay protocol
 in x-by-wire applications such as drive-by-wire, steer-by-wire and break-by-wire.

Tracking using square region

Tracking using active contour region

source: http://www.adandp.media/articles/a-manager's-guide-to-flexray


Home

Amir Rasouli, M.A.Sc
PhD Candidate at YorkU


Home


Research


CV


Code


Publications


Contact

Swarm of Interacting Reinforcement Learners (SWIRLs)











     



     



     


     This work is an implementation of the paper, "Distributed Path Planning for Mobile Robots using a  
Swarm of Interacting Reinforcement Learners", by M. Vigorito in Java. The implementation includes various levels of concurrency (e.g. sequential vs multi threaded).  code    paper

Fault-Tolerant Gathering for Autonomous Mobile Robot















     This is a simulation of the gathering algorithm for a swarm of autonomous mobile robots presented in the paper, " Fault-Tolerant Algorithms for Autonomous Mobile Robots" by Agmon and Peleg. The code is implemented in C++ and demonstrates how a group of robots can converge to a point despite the presence of a number of faulty (Byzantine) robots. code     slides 

A Visual Solver for the Hexabits Puzzle













    This is a visual puzzle solver for the Hexabits game. The code is in MATLAB and uses a puzzle tile generator to create the game, a visual recognition to identify their patterns and a solver to place the tiles in correct places. code     report


A Command Line Demo for Cyton Gamma 1500









     This is a sample code for using robotic arm, Cyton Gamma 1500 from Robai. The code is written in C++ and contains a client and a server ROS packages to be used in any application. It also include a demo code that allows the user to manipulate the arm in command line using keyboard inputs.  code

Gazebo World Plugin for Model Manipulation






     A world plugin for manipulating objects and light sources in Gazebo world environments. The code is written in C++ and uses ROS.    code

A Saliency Framework for Visual Search

     










     This is the implementation of the saliency model used in my paper, "Visual Saliency Improves Autonomous Visual Search". The code is in C++ and wrapped as a ROS package.  code

A multi-threaded implementation of SWIRLs

Amir Rasouli, M.A.Sc
PhD Candidate at YorkU


Home


Research


CV


Code


Publications


Contact

Publications:
  • A. Rasouli, I. Kotseruba and J. K. Tsotsos (2017), Agreeing to Cross: How Drivers and Pedestrians Communicate, Intelligent Vehicle Symposium (IV), Redondo Beach, Jun. arxiv link
  • I. Kotseruba, A. Rasouli and J. K. Tsotsos (2016), Joint Attention in Autonomous Driving (JAAD), arXiv:1609.04741, Sep. link
  • A. Rasouli and J. K. Tsotsos (2016), Sensor Planning for 3D Visual Search with Task Constraints, in Proc.  Canadian Conference on Computer and Robot Vision, Victoria, May.  link
  • A. Rasouli and J. K. Tsotsos (2015), Integrating Three Mechanisms of Visual Attention for Active Visual Search, in Proc. 8 The International Symposium on Attention in Cognitive Systems IROS, Hamburg, Oct.  paper
  • A. Rasouli (2015), Attention and Sensor Planning in Autonomous Robotic Visual Search, Master’s Thesis, Jan.  link
  • A. Rasouli and J. K. Tsotsos (2014), "Attention in Autonomous Robotic Visual Search", in Proc. i-SAIRAS, Jun.  paper
  • A. Rasouli and J. K. Tsotsos (2014), "Visual Saliency Improves Autonomous Visual Search", in Proc. The Conference on Computer and Robotic Vision, May.  link
Posters:
  • A. Rasouli, M. Solbach and J. K. Tsotsos (2016), "View-Invariant Active Visual Search", in Proc. NCFRN, Jun.  poster
  • A. Rasouli, Iullia Kotseruba and J. K. Tsotsos (2015), "Visual Saliency in Search and Exploration of Unknown Environments",  in Proc. NCFRN, Jun.  poster
  • A.Rasouli and J. K. Tsotsos (2014), "Saliency as Look Ahead in Robotic Visual Search", in Proc. NCFRN, May.  poster
  • C. Wloka, T. Fischer, E. Simine, A. Rasouli, J. P. Gignac, D. Wick, S. Sutherland and J. K. Tsotsos (2013), "VirtualMe: The Telepresence Robot", in Proc. NCFRN, May.  poster
  • A. Rasouli and Paul Beckett (2010), "FlexRay Network Solutions for Control & Safety in Automobiles", Jun.
Datasets:
  • 3D Gazebo Models (3DGEMS): a dataset of 270+ 3D models and world files designed for Gazebo simulation software. link
  • Joint Attention in Autonomous Driving (JAAD): a dataset of 340+ clips of traffic scenes for the purpose of analyzing pedestrians' behaviors at the point of crossing. link 

Amir Rasouli, M.A.Sc
PhD Candidate at YorkU


Home


Research


CV


Publications


Contact


Code

Address:
  • Room 3054, Dept. Electrical Engineering and Computer Science, Lassonde School of Engineering, Lassonde Bldg. 120 Campus Walk, York University, 4700 Kelee St., Toronto, Ontario, Canada M3J 1P3, map
Tel:
  • +1-416-736-2100 ext. 33972

Email:
  • aras@eecs.yorku.ca

Academic Affiliations:
  • Prof. Tsotsos Active and Attentive Vision, link
  • Lassonde School of Engineering, link
  • York University, link
  • NSERC Canadian Field Robotic Network (NCFRN), link
  • The Center for Vision Research, link
Some Websites of Interest:
  • Autonomous Driving New, link
  • IEEE Spectrum Robotic News, link