Proficiency

• Specialized in computer vision, image processing, machine learning,and pattern recognition.
• Skilled in employing MATLAB for prototyping and implementing algorithms.
• Proficient in C/ C programming using object-oriented analysis and design approach.
• Excellent in code optimization restructuring dataflow, SIMD, removing bottlenecks in pipeline, and fixed point design.
• Adept in statistics and dealing with large data sets for statistical modeling.
• Good analytical and problem solving skills, team player with the ability to work individually also, quick learner, and adaptive to latest technologies.

Work Experience

Software Developer

Confidential

• Developed an e-fingerprint product that locates and decodes barcodes, extracts e-fingerprint from it, and saves it in a database in its training phase on a manufacturing plant. In the test phase, an iOS app sends a captured image to a server running the authentication module. This module extracts e-fingerprint from the barcode, compares it to the stored e-fingerprint if available, and sends a reply to the phone with a classification.
• Designed and developed an e-fingerprint C shared library to locate and decode multiple Data matrix and QR code and extract e-fingerprint from them. This was done using STL/OpenCV on windows platform. These libraries were also made compatible for Linux platforms.
• Designed a test setup in C for regression testing in terms of speed, accuracy, and sustainability of the C library.
• Developed an authentication module using the e-fingerprint technology using C and STL. The authentication module interface was made compatible for Python-based web applications.
• Developed a test setup to mimic the test phase in Python using Request library which will push images through the HTTP post and receive a reply from the authentication module running on the server. The script sporadically and randomly pushes images from a set to test the processing capacity during simultaneous requests to the server.
• Employed the preliminary design using sockets TCP/IP for data transfer within processes on Windows and Linux.
• Implemented a future data flow process module using RabbitMQ, a messaging system, to transfer the e-fingerprint from the manufacturing plant to the database. This module is under testing.
• Developed and executed a barcode detection module which involved reviewing segmentation, region growing approaches, and blob analysis. It also included checking for the presence of a grid of lines detected from edges,the purpose of which was locating barcodes in cluttered surroundings.
• Generated a feature pool which included investigation of position of deviation from line fit, discrepancies within modules, and curvature analysis on contours.
• Confidential
• Designed and developed a demographics detection module that performs gender classification, age estimation, and expression recognition by reading facial images using C , Boost, STL, and OpenCV.
• Designed and developed a face detection module for audience analytics that involves feature generation and building classifier using C , OpenCV, and STL.
• Collected and generated metadata dataset for training and testing of demographic classifiers.
• Designed and implemented the prediction engine for adaptively changing advertisements based on demographics of the audience. This module tracked the frequency of the advertisements displayed and avoided repetitive broadcast as and when required.
• Investigated features like Haar, LBP, and Gabor prior to development. Prediction and classification involved usage and implementation of algorithms such as SVM, boosting cascades, and classification trees in MATLAB and C .
• Divided the dataset into train, validate, and test sections for optimization of accuracy vs. speed.

Confidential Executed pedestrian detection algorithm using HOG features, based on the Open CV People Detects code. Improved the execution time of the system by 130X using algorithm restructuring and kernel based optimization. Designed and optimized a fixed point image rescaling function using bilinear interpolation for TI's vision library VLIB .

Confidential

Knowledge Development and Application

• Functionality: This image processor will serve as a tool for image editing and image analysis techniques.
• Development Framework:
• C for a graphical user interface
• EmguCV .Net port of OpenCV for processing images.
• Currently the Image Processor:
• Thresholds images globally and locally using OSTU's method or a user defined value.
• Converts images from color to gray scale or saves images into various formats.
• Warps images according to a given set of points using affine and perspective warping.
• Graphically selects the set of points for warping and a rectangle for cropping images.

Data analysis

• Data Source:Kaggle or free secondary data sets.
• Statistical Tool: R
• Statistical Technique:Predictive modeling over a population or classification.

Skills

Programming Languages: C, C , .NET/C , Python, MATLAB, Octave, SAS, MySQL

Statistical Tools: SAS Enterprise Miner, SAS Enterprise Guide, R

Embedded Systems: VLIW DSP architectures - TI C6x and C674x, SIMD TI DSP

Libraries: OpenCV, STL, IPP, CMake, RabbitMQ

Development Tools: Visual Studio, Code Blocks, Mercurial Bitbucket