Anup Doshi

Graduate Student Researcher
Computer Vision and Robotics Research Laboratory
University of California, San Diego
La Jolla, CA 92037
andoshi_at_ucsd.edu

Current Research
Prior Research
Education and Honors

Current Interests and Projects at CVRR

Driver Assistance Systems in Intelligent Automobiles
With the LISA-P Intelligent Testbed as a powerful platform for researching driver assistance systems, we are developing and analyzing vision-based methods to infer driver behavior and responses. Current experiments include evaluating driver response to visual feedback based on the driver's behavior. We are also interested in developing machine-learning based methods for driver pose analysis and intent prediction.

Publications

  • Anup Doshi, Mohan M. Trivedi,  A Comparative Exploration of Eye Gaze and Head
    Pose Cues for Lane Change Intent Prediction, IEEE Intelligent Vehicles Symposium, June 2008.
  • Anup Doshi, Shinko Y. Cheng, Mohan M. Trivedi, A Novel, Active Heads-up Display for Driver Assistance, IEEE Transactions SMC-B Special Issue on Human Computing, 2008.
  • Erik Murphy-Chutorian, Anup Doshi, Mohan M. Trivedi, Head Pose Estimation for Driver Assistance Systems: A Robust Algorithm and Experimental Evaluation, In IEEE Conf. Intelligent Transportation Systems, Sept 2007. (pdf)
  • Shinko Y. Cheng, Anup Doshi, Mohan M. Trivedi, Active Heads-up Display based Speed Compliance Aid for Driver Assistance: A Novel Interface and Comparative Experimental Studies, IEEE Intelligent Vehicles Symposium, June 2007. (pdf)

    • Robust Background Subtraction and Shadow Suppression for Outdoor Surveillance
    Motivated by the need for deployable and robust real-time background subtraction methods in outdoor test-beds, we have been developing techniques for reliable foreground segmentation. The Hybrid-Cone-Cylinder Codebook model incorporates methods for adaptive background modeling and shadow suppression into a sensitive and coherent methodology. Additionally, we have examined the use of satellite imagery as a multi-sensory source of data to inform illumination parameters in a cloudy scene.

    Publications

  • Anup Doshi, Mohan M. Trivedi, Satellite Imagery Based Robust, Adaptive Background Models and Shadow Suppression. Signal, Image, and Video Processing (SIViP) Journal, 1(2):119-132, June 2007 . (pdf)
  • Anup Doshi, Mohan M. Trivedi, Hybrid Cone-Cylinder Codebook Model for Foreground Detection with Shadow and Highlight Suppression. Proc. IEEE International Conference on Advanced Video and Signal based Surveillance, Nov 2006. (pdf)

    • Other Research @ UCSD

    A Mind-reading Machine
    Under the guidance of Prof. Yoav Freund, an applet was developed to test the abilities of a machine-learning based 'Mind-reading' applet and collect data to enhance its performance. The applet is based upon active learning methods from "A Context-Tree Weighting Method" [Willems et al.] and more info can be found here.

    People Counting and Tracking for Surveillance
    The goal of this project is to implement a system to count and track the people in a database of surveillance footage. Due to several constraints and performance issues, a straightforward algorithm based on background subtraction is implemented and shows acceptable performance levels. Further improvements are considered to improve the performance, including implementations of algorithms such as BraMBLe.

    Undergrad Research Projects @ Carnegie Mellon

    3D Image Processing - Senior Research Project
    The goal was to synthesize 3-D models of human faces from video camera sequences. With depth information from the incoming sequences, I built a 3-D model of the subject's face. Advisor: Dr. Tsuhan Chen. Grad Students: Todd Stephenson.

    Time Reversal - Senior Honors Research Project
    Helped develop code for a project using time reversal techniques in a MIMO system to characterize a communications channel. Automated the measurements process by integrating various motors and the network analyzer using C++. Advisor: Dr. Dan Stancil. Grad Students: Ahmet Cepni, Ben Henty.

    Microphone Array Processing - 18551 Course Project
    Spring 2004 - Implemented a phased-array of microphones that can localize and track the source of a sound in real time, amidst noise. Led a group of 4 in planning, design, and programming on a TI TMS320C67 DSP. Optimized the algorithm using MATLAB, C and Assembly.

    Superresolution - 18798 Course Project
    Spring 2004 - Designed and implemented a fast algorithm for Super-resolution enhancement of video sequences in MATLAB and C. Tried to updated the implementation to use Optical Flow for sub-pixel accuracy.

    UWB Interference Effects - Clemson University SURE REU
    Summer 2003 - Conducted a research project on properties of Ultra-wideband Communications. Modeled and analyzed UWB signals and interference on narrowband devices. Implemented a Monte Carlo simulation of UWB interference effects on 'victim' receivers. Advisor: Dr. Carl Baum

    Education and Honors

  • University of California, San Diego, La Jolla, CA
        Candidate for Ph.D., Electrical and Computer Engineering
        Specialization: Signal and Image Processing. GPA: 4.0 / 4.0
  • Carnegie Mellon University, Pittsburgh, PA
        M.S., Electrical and Computer Engineering, May 2005. GPA: 4.0 / 4.0
        B.S., Electrical and Computer Engineering, May 2005. GPA: 3.96 / 4.0

    Honors
    - Jacobs Fellow, University of California, San Diego, 2005-2008
    - Finalist, Hertz Foundation Graduate Fellowship, 2005
    - Honorable Mention, NSF Graduate Fellowship, 2005
    - IEC-William L. Everitt Student Award of Excellence, Carnegie Mellon, 2005
    - Andrew Carnegie Merit Scholarship, Carnegie Mellon, 2001
    - Member, Eta Kappa Nu, Tau Beta Pi, and Phi Kappa Phi Honor Societies

    • Home