PROJECTS

(This page contains the description of some of my older projects, dating back to my Masters and undergraduate academics!)

M.Sc. thesis [Link]
Visual Question Answering with Modules and Language Modeling
Thesis committee: Alain Tapp, Christopher J. Pal and Yoshua Bengio

• A mixture of experts formulation to sparsely execute the experts in a convolutional block.
• Learning the internal structure of modules in terms of elementary arithmetic operations for Neural Module Networks.
• Learning a language model which makes explicit use of intermediate parse structure and can be trained in an unsupervised manner.
  

B.Tech Project   [Thesis report]
Inference of Probabilistic Regulatory Networks (PBN) for Gene Regulation
Advisor: Dr. R.K. Layek and Dr. Pabitra Mitra
July 2015 - April 2016

• Proposed a novel algorithm to infer a PBN for gene regulation using biological pathways and time-series data.
• Application of above algorithm to infer a robust PBN for modelling the yeast cellcycle trajectory.
  

Research Internship at Xerox Research Centre India   [Internship Report]
Project Title: Object Recognition and Activity Classification of Video Activities
May 2015 – July 2015 | Bangalore, India
Guide: Dr. Arijit Biswas

• Trained Convolutional Neural Network (CNN) for egocentric videos’ object detection.
  
• Activity classification using Temporal Pyramid based features through Support Vector Machines.
  
• Used CNN features to train Long Short-Term Memory (LSTM) model for activity classification.
  

Texas Instruments Innovation Challenge(Team Project)
Reading Assistance for Blind/Low - Vision People
Jan 2015 – Feb 2015 | IIT Kharagpur
Guide: Dr. Gautam Das

• Designed a text-to-speech converter prototype using Beagle Bone which takes images of book pages and reads the text aloud including a mechanism for book page turning.
  
• Implemented camera image pre-processing including perspective correction for OCR using Tesseract OCR engine.
  [Report] [Demo]
  

Thermal Aware Mapping Algorithm for Network-on-Chip
Guide: Prof. Santanu Chattopadhyay
Dec 2014 | IIT Kharagpur

• Implementation of CoolMap: a temperature-aware mapping algorithm to minimise thermal correlation and communication cost between IPCores.
  [Report]
  

Real-Time Drivable Road Detection
Guide: Prof. Manjunatha Mahadevappa
June 2014 – July 2014 | IIT Kharagpur

• Implemented Road Detection using Neural Network employing Saliency Maps as neural network input feature in OpenCV.
  
• Implemented Real-time Drivable Road Detection and Obstacle Detection using Region growing Segmentation.
  [Report]