Areas of Interest
- Computer Vision
- Image Processing
- Machine Learning
Courses
- Graduate: Image Understanding, Computational Linguistics 2, Statistical Pattern Recognition, Convex Optimization, Sparse Statistical Signal Processing & Learning, Digital Image & Video Processing, Estimation & Detection Theory, Random Processes in Communication & Control, Advanced Digital Signal Processing, Mathematical Foundations for Computer Engineering
- Undergraduate: Digital Image Processing, Digital Signal Processing, Soft Computing Tools in Engineering, Artificial Intelligence, Data Communication & Networks, Instrumentation System Design, Database Management Systems, Process Monitoring & Fault Diagnosis, Process Dynamics & Control, Instrumentation Devices
- MOOC: Introduction to Big Data with Apache Spark, Scalable Machine Learning, Machine Learning, Web Intelligence & Big Data, Computer Networks, Heterogeneous Parallel Programming, Discrete Inference & Learning in Artificial Vision, Data Analysis & Statistical Inference, Control of Mobile Robots
Medifor
Currently working on a project which seeks to detect if people present in an image are doctored or digitally added. Developing algorithms to get illumination parameters from a given 2D image and to use illumination discrepancies to detect tampering.
Active Authentication
Worked in a DARPA funded project to develop methods to detect if a mobile device is being used by unauthorized users. Primarily focussed on developing face detection algorithms tailored for the mobile domain.
Publications include:
- S. Sarkar, V. M. Patel, and R. Chellappa, “Deep feature-based face detection on mobile devices,” in IEEE International Conference on Identity, Security and Behavior Analysis, 2016.
- U. Mahbub, S. Sarkar, V. M. Patel, and R. Chellappa, “Active User Authentication for Smartphones: A Challenge Data Set and Benchmark Results” in IEEE International Conference on Biometrics: Theory, Applications, and Systems, 2016. Won best poster award.
- U. Mahbub*, S. Sarkar*, V. M. Patel, and R. Chellappa, "Pooling Facial Segments to Face: The Shallow and Deep Ends" in IEEE Conference on Automatic Face and Gesture Recognition, 2017.
- U. Mahbub*, S. Sarkar*, and R. Chellappa, "Segment-based Methods for Facial Attribute Detection from Partial Faces" in IEEE Transactions on Affective Computing.
- U. Mahbub, S. Sarkar, and R. Chellappa, "Partial Face Detection in the Mobile Domain" submitted to Image and Vision Computing.
- S. Sarkar, A. Bansal, U. Mahbub, and R. Chellappa, "UPSET and ANGRI : Breaking High Performance Image Classifiers"
* mean equal contributions
Signal and Image Processing
- Image stitching and inpainting
- Implemented image stitching by SIFT based keypoint detection, RANSAC and Laplacian Pyramid blending.
- Developed exemplar based inpainting applications using OpenCV for Windows and Android
- Foreground extraction in surveillance videos using sparsity based methods
- Studied multiple algorithms like RPCA, block sparse RPCA, SpaRCS among others that use sparsity based methods to separate the static (low rank) background and the dynamic (sparse) foreground, which was then used to detect anomalies.
- ECG signal acquisition, feature extraction, classification and indexing
- Designed and simulated the analog frontend for ECG signal acquisition
- Extracted relevant features from the signal using curve length transform (using MATLAB) and slope based method (using MATLAB and C55x Digital Signal Processors emulator) and then used them to linearly classify ECGs into healthy and high-risk clusters
- Explored kd-trees and hash tables to index a multidimensional ECG database.
Convex Optimization
- Cvx4py: A cvx like wrapper for solving convex problems in Python
- Developed a framework, cvx4py, that lets one use cvx code written for MATLAB, directly in Python, to solve Convex Optimization problems. Second Order Conic Programs, Geometric Programs and Semi Definite Programs are supported.
Machine Learning
- Detecting schizophrenia through tweets
- Cleaned, analysed and extracted features from a tweet dataset to train classifiers to detect schizophrenic tendencies.
- Features that captures schizophrenic behaviour, like presence of rhymes and complexity of sentences were proposed.
- Relation Predicting Algorithms in Social Networks
- Trained a Logistic Regression Classifier using features from the social graph of Epinions users to predict the relation between two users given their relations with other users. Two models of 7 and 16 features were studied and compared.
Computational Software
- Optimal order of Gaussian Elimination for sparse symmetric matrices
- Represented sparse matrices as graphs and determined lexicographic ordering for vertex elimination
- Using the order found, performed Gaussian Elimination to solve the linear equations represented by the matrix.
- Equisolve, an application for solving simultaneous Algebraic Equations
- Used soft computing algorithms like Differential Evolution (DE) and Genetic Algorithm (GA) to solve systems of general Simultaneous Algebraic Equations and developed a MATLAB based GUI application, ‘Equisolve’
Embedded Systems
- Modelling of Resistive Touchpad
- Studied touchpad technologies currently available especially the resistive touchpad technology and developed a resistive mesh connected with an ATmega32 microcontroller and an LCD display to display coordinates of the contacted point