Alireza Fathi


Ph.D. Student, Wall Lab, School of Interactive Computing

College of Computing
Georgia Institute of Technology

Email:  afathi3 at gatech dot edu

Research | Services | Courses | Useful Links | Experience | Fun! | My Sister
 

Formal Bio: I am currently a Ph.D. student at GeorgiaTech working with Jim Rehg. I worked with Greg Mori during my Masters. I have closely worked with Jessica Hodgins, Frank Dellaert and John Krumm also in the past.

Informal Bio: I was born at 1984 in Tehran, Iran. I really love my country and I miss it so much. I am proud of my country's art, old history, and its kind and lovely nation. I am also missing Iranian Food so much. Get some food recipes here. I can cook Halva, Sho'le zard, Fereni, Fesenjoon, Kookoo Sabzi, Abgoosht, Ghormesabzi. The food and dessert recipes I wish to learn are Ashe reshte, Beryani, Naan Berenji, Baghlava, Ghottab.

 

News

05/03/2012: I proposed on May 3rd, 2012. My thesis committee members are James Rehg, Martial Hebert, Antonio Torralba, Gregory Abowd, Irfan Essa and Thad Starner.

02/26/2012: Paper accepted to CVPR 2012, Social Interactions: A First-Person Perspective (PDF), get the dataset here.

01/21/2012: I am the co-chair of 2nd IEEE workshop on Egocentric (First-Person) Vision in conjunction with CVPR 2012.

11/08/2011: I am a program committee member of AAAI 2012.


06/28/2011: Paper accepted to BMVC 2011: Combining Self Training and Active Learning for Video Segmentation (PDF, Abstract).

06/10/2011: Paper accepted to ICCV 2011: Understanding Egocentric Activities. PDF

02/14/2011: GeorgiaTech Egocentric Activities Dataset is released, get it here.

02/14/2011: Paper accepted to CVPR 2011: Learning to Recognize Objects in Egocentric Activities. PDF

10/28/2010: GTSAM is released. Being part of Frank Dellaert's group for a year, all of his students including myself and Frank developed a library of C++ classes that implement smoothing and mapping (SAM) in robotics and vision, using factor graphs and bayes networks (Link).

03/15/2010: GIScience paper accepted (PDF).

 
Research
My research is in the area of Computer Vision, Machine Learning and Robotics. I am particularly interested in the following topics:

  • Egocentric (First-Person) Vision (CVPR11, ICCV11, CVPR12): An egocentric vision system, is a framework consisting of a wearable camera that continuoulsy captures the scene in front of the first-person. In particular, I define an egocentric vision system as a framework that leverages different levels of first-person attention to identify important objects and faces in the scene that contribute to subject's activities. First-person's attitude, including where she looks (gaze) and what she does (hands manipulating objects) provide an invaluable context for determining the objects that grab her attention at any given time. Our goal is to use these structured sources of information coming from first-person in order to enable weakly supervised recognition of objects and activities.

  • Action Recognition (ICCV07, CVPR08, ICCV11, CVPR12, MSc Thesis): I aim at developing action recognition techniques that rely on semantically meaningful features which capture interaction of objects with each other. This is in contrast to state of the art techniques that are based on space-time interest points or point trajectories.

  • Semi-Supervised and Active Learning (BMVC11, CVPR11): Today, it is a well known fact in computer vision community that the amount of visual data in the world is significantly more than the available annotation resources. This has created a big demand for semi-supervised learning and more recently active learning techniques in vision. So far, web has been the main source of visual data for computer vision researchers. However, I believe, Egocentric vision, will be an alternative in the near future. High-quality, tiny and inexpensive wearable cameras are becoming publically available, and every day thousands of hours of egocentric videos are being recorded. Egocentric videos have various advantages in comparison to typical videos on the web: (1) egocentric videos are well structured and are associated with context provided by the first-person (first-persos's voice, first-person's hands, gaze, etc.), (2) the objects and scenes in egocentric videos are what humans observe throughout their lives, as a result the egocentric data is less biased in comparison to web data and (3) the egocentric videos are captured from human point of view.

  • Video and Image Segmentation (BMVC11): I believe that segmentation is probably the most fundamental problem in computer vision. If segmentation is solved, many of the big challenges in the field become trivial.

  • Localization and Mapping (GIScience10, Tech10-1, Tech10-2): Unfortunately, there are very few researchers in the field, who work on inferring the 3d structure of the world, and use it for understanding videos and images. I have separately worked on SLAM and 3D reconstruction, and on the other hand activity recognition and video segmentation. However, these are not disjoint problems, and the information from each one can significantly improve the other one. I like to further focus on combining the 3d reasoning with other tasks in computer vision.

 
Egocentric (First-Person) Vision

  • Alireza Fathi, Xiaofeng Ren, James M. Rehg, Learning to Recognize Objects in Egocentric Activities, CVPR, 2011. (PDF, Dataset)

  • Alireza Fathi, Ali Farhadi, James M. Rehg, Understanding Egocentric Activities, ICCV, 2011. (PDF, Dataset)

  • Alireza Fathi, Jessica K. Hodgins, James M. Rehg, Social Interactions: A First-Person Perspective, CVPR, 2012. (PDF, Dataset)



     
    Video Segmentation

  • Alireza Fathi, Maria Florina Balcan, Xiaofeng Ren, James M. Rehg, Combining Self Training and Active Learning for Video Segmentation, BMVC, 2011 (PDF, Abstract).



    Action Recognition

  • Alireza Fathi, Greg Mori, Action Recognition by Learning Mid-level Motion Features, CVPR, 2008. (PDF, Bibtex)

     
    Human Tracking

  • Alireza Fathi and Greg Mori, Human Pose Estimation using Motion Exemplars, ICCV, 2007. (PDF, Bibtex, More Information, Slides, Course Project that led to this paper)

  • MSc Thesis: Alireza Fathi, Human Figure Tracking using Motion Exemplars, Department of Computing Science, Simon Fraser University, 2008. (PDF)
     
     
     
    Localization and Mapping

  • Helped in developing GTSAM as part of Frank Dellaert's team.

  • Alireza Fathi, John Krumm, Detecting Road Intersections from GPS Traces, GIScience, 2010. (PDF)

  • Alireza Fathi, Alex Cunninghum, Balmanohar Paluri, Kai Ni and Frank Dellaert, EasySLAM, GVU Technical Report(GIT-GVU-10-03), 2010. (Link)

  • Frank Dellaert, Alireza Fathi, Alex Cunninghum, Balmanohar Paluri and Kai Ni, Local Exponential Maps: Towards Massively Distributed Multi-robot Mapping, GVU Technical Report(GIT-GVU-10-04), 2010. (Link)
     

     

     
    Color Constancy
     
  • Mark S. Drew, Muntaseer Salahuddin, Alireza Fathi, A Standardized Workflow for Illumination-Invariant Image Extraction, 15th Color Imaging Conference, New Mexico, 2007. (PDF)
     
    Robocup
     
  • Nasrin Mostafazadeh, Saba Ardeshiri, Sepideh Movaghati, Shadi Hariri, Zeinab Jahanzad, Alireza Fathi, Majid Valipour, Poseidon Team Description Paper, RoboCup 2006, Bremen, Germany. (PDF)

  • Saman Aliari Zonouz, Hamid Reza Vaezi Joze, Siavash Rahbar, Majid Valipour, Alireza Fathi, Impossibles Sony Aibo 4-Legged RoboCup Technical report, RoboCup 2006, Bremen, Germany. (PDF)

  • Hamid Reza Vaezi Joze, Saman Aliari Zonouz, Siavash Rahbar, Majid Valipour, Alireza Fathi, Impossibles Sony Aibo 4-Legged Team Description Paper, RoboCup 2006, Bremen, Germany. (PDF)

  • Jafar Habibi, Alireza Fathi, Saeed Hassanpour, Mohammad Reza Ghodsi, Behzad Sadjadi, Hamid Reza Vaezi, Majid Valipour, Impossibles Team Description Paper, RoboCup 2005, Osaka, Japan. (PDF)

     

     
    Miscellaneous

     
  • Tamara Smyth, Alireza Fathi, Voice Synthesis using the Generalized Pressure-Controlled Valve, International Computer Music Conference (ICMC), 2008. (PDF)
     
  • BSc Thesis: Alireza Fathi, Assembler and Simulator for IBM 360/370, Computer Engineering Department, Sharif University of Technology, 2004. (Director: Dr. Hamid Sarbazi Azad) (Persian PDF)

     

     

     

    Courses

     

    Services
     
    Co-Chair of 2nd IEEE Workshop on Egocentric (First-Person) Vision in Conjunction with CVPR 2012
    Program Committee member of AAAI 2012.
    Reviewer of IEEE Transaction on Pattern Analysis and Machine Intelligence (PAMI).
    Reviewer of IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
    Reviewer of International Conference on Computer Vision (ICCV).
    Reviewer of British Machine Vision Conference (BMVC).
    Reviewer of ACM Conference on Ubiquitous Computing (UbiComp).
    Reviewer of IEEE Transactions on Circuits and Systems for Video Technology (TCSVT).
     

    Useful Links

  • Notes on Graduate Studies, Alireza Fathi, 2010.
  • How to Buy a Used Car, Alireza Fathi, 2010.
  • Recent Hot Machine Learning Hammers used in Computer Vision, Alireza Fathi, 2011.
  • My Facebook App for Data Collection
  • Simultaneous Recovery of Shape, Motion and Grouping by Applying Rank Constraints, 2010. (PDF, Poster)



    Experience

    Teaching Experience:

    Research Experience:

    Work Experience:

    Languages:

     

    Fun

    This video is seen by 20,000 people by now (Nov 2010)

     

     

    website statistics