Project Computer Science

Prithviraj Banerjee

**** **** ***, *** *** Phone : +1-323-***-****

University of Southern California E-mail : ********.**********@*****.***

Information

Los Angeles, CA-90089 Web : http://www.prithviraj.net

Computer Vision, Human Computer Interaction, Machine Learning, Pattern Recognition,

Research

Dynamical models for activity perception and recognition.

Interests

University of Southern California, Los Angeles, California, USA

Education

Doctor of Philosophy, Computer Science August 2008 present

Expected graduation date: May 2013

Advisor: Professor Ram Nevatia

M.S., Computer Science August 2008 May 2011

GPA: 4.0 out of 4.0

Advisor: Professor Ram Nevatia

Indian Institute of Technology (IIT), Kharagpur, West Bengal, India

M.Tech. & B.Tech., Electronics and Electrical Communication August 2003 May 2008

GPA: 9.25 out of 10.00

Master s Specialization: Visual information and Embedded Systems

Thesis Topic: Framework for Human Tracking and Model Generation

Advisor: Professor Somnath Sengupta

Provost Fellowship, University of Southern California, 2008-2010. Awarded to the top 100 Ph.D.

Honours and

students across all disciplines.

Awards

Keshab Parhi Award for the best demonstrable Master s thesis project of practical interest, Indian

Institute of Technology, Kharagpur, 2008

Scholarship for interning at the School of Informatics at University of Edinburgh, 2006 (declined).

P. Banerjee and R. Nevatia, Pose Based Activity Recognition using Multiple Kernel Learning,

Publications

IAPR International Conference on Pattern Recognition (ICPR), 2012, Japan.

P. Banerjee and R. Nevatia, Learning Neighborhood Co-occurrence Statistics of Sparse Features for

Human Activity Recognition, IEEE Advanced Video and Signal-Based Surveillance (AVSS), 2011,

Austria.

P. Banerjee and R. Nevatia, Dynamics based Trajectory Segmentation for UAV videos, IEEE

Advanced Video and Signal-Based Surveillance (AVSS), 2010, Boston, USA.

P. Natarajan, P. Banerjee and R. Nevatia, Accurate Person Tracking Through Changing Poses for

Multi-view Action Recognition, ICVGIP, 2010, India.

P. Natarajan, P. Banerjee, F. Khan and R. Nevatia, Graphical Framework for Action Recognition

using Temporally Dense STIPs, IEEE WMVC, 2009, Utah, USA.

P. Banerjee, A. Pinz and S. Sengupta, Model generation for robust object tracking based on Tem-

porally Stable Regions, IEEE WMVC, 2008, Colorado, USA.

P. Banerjee and S. Sengupta, Human Motion Detection and Tracking for Video Surveillance,

National Conference for Communication, 2008, Bombay, India.

Patent application (Application Number: 200*-******) titled Operation Analysis Apparatus led

for publication in Japanese Translations of PCT International Publication for Patent Applications in

2006, with co-inventors T. Hashimoto, T. Katooka, M. Haruda and A. Constancin

University of Southern California, Los Angeles, California USA

Research

Experience

Graduate Student August 2008 present

Human Activity Analysis: Evaluation of spatio-temporal interest point (STIP) features for human

activity recognition. Unstructured STIP-based features were mapped to a structured graphical

model representation, that allows continuous action recognition on cluttered scenes with scale and

viewpoint variations. Hidden Conditional Random Fields are proposed for modeling the pairwise

local neighborhood relationships of STIP feature detections in a video volume. A parallel line of

work has been on exploring structured activity models based on key poses of human activity.

UAV Video Analysis: Worked on event recognition in vehicle trajectories from UAV videos

(DARPA-VIRAT project), and exploring how modeling the dynamics of a vehicle can be helpful

in event recognition. Currently working on noise tolerance analysis of hierarchical graphical

models for recognizing complex activities in UAV videos.

Indian Institute of Technology (IIT), Kharagpur, India

Undergraduate Student August 2003 May 2008

Human Detection and Tracking System for Video Surveillance. The work involved the tracking

of objects in motion and classifying them as Human or Non-Human entities using a Histogram

of Oriented Gradients based human detector.

Model generation for Robust Object Tracking (Master s thesis). A Temporally Stable Region

(TSR) based object tracking system was developed with focus on tracking objects through par-

tial and complete occlusion (received the Keshab Parhi Award).

Institute of Electrical Measurement and Measurement Signal Processing, TU Graz, Austria

Internship May 2007 July 2007

A new approach to tracking objects was implemented through the generation of Temporally Stable

Regions and graph based matching.

Software R&D Group - Ricoh Company, Tokyo, Japan

Internship May 2006 July 2006

The work involved temporally segmenting the video of a factory worker assembling a product into

key steps of the assembly operation and automatically extracting the various characteristic traits of

the worker (Patent Application No.: 200*-****** led).

Division for Robotics & Remote Handling, Bhabha Atomic Research Centre (BARC), India

Internship May 2005 July 2005

The project involved setting up a Serial Communication interface between a remote robot and the

base station, primarily for picture and range data transmission, creating GUIs for the Sensor Data

Acquisition model and documenting existing codes for future reference of the lab.

C, C++, Java, Matlab, SQL.

Programming

Computer Vision Machine Learning Analysis of Algorithms

Graduate

Probability for Engineers Probabilistic Reasoning Optimization: Theory & Algorithms

Courses

Database Systems Computer Communications Non-Linear Prog. for Optimization

Provided on request.

Referees

Contact this candidate