Research & Experience

Vehicle OverSpeed Detection and Recognition
This project started as a final year thesis project. By the end of this project we created a Doppler radar speed detector and a number plate recogniser. The concept of this project is to detect overspeeding vehicles in the road and automatically register their number plate using Image Processing Techniques. The project was designed in an attempt to assist Traffic Police.

Automatic Number Plate Recognition With Support Vector Machine
This is an independent research focused on creating a robust and accurate License Plate Recognition system. The developed system applies color based segmentation to isolate LIcense Plate and its characters. Recognition of the characters is done using support vector machines.

Devanagari Handwritten Character Dataset(DHCD)
By the end of the DHCD project we created the largest image database for handwritten Devanagari characters. This research was focused on helping the researchers working in optical character recognition of Nepali, Hindi and other languages derived from Devanagari script. The Dataset consists of 92 thousand unique images of 46 different characters. The dataset is divided into training set and test set. The dataset is publicly available for research purpose.

Deep Learning Based Large Scale Handwritten Devanagari Character Recognition
Deep Neural Networks have shown great results in many pattern recognition tasks. So the optical character recognition for Devanagari characters was designed and developed using Deep convolutional Neural Network. We have proposed two different architectures of Deep CNN for recognition. We have also implemented overfitting prevention technique. The proposed architecture obtained highest test accuracy of 98.4 % in the Test set in DHCD.

Face Detection and Verification System
Face Detection and Verification is a very important system especially for Security purposes. I have experience of working with recognition and verification system for more than a year. I have worked to create very large image datasets for Face detection and recognition. I have implemented Deep Face , Deep ID, GoogleNet, Inception, VGG, siamese and other similar models for face recognition and verification. Beside this, I have also worked in Face Tracking and Object Detection.

Beautiful Mind Project
This project was completed in collaboration with UNICEF Nepal and Autism Care Nepal Society. We have developed an Android Application that is helpful to autistic children and their parent. We expect the application to bridge the communication gap between autistic children and their parents while also helping us to better understand the social behavious of autistic children.

Beside these, I have previously worked in a number of projects. I have experience of designing and implementing product recommendation systems and friend suggestion systems. I have also worked on Natural Language Processing projects especially for time and event parsing, POS tagging and Named Entity Recognition. I have worked to implement Google Map API services in different projects, helped develop offline map-service using OSRM and implement cross-language RPC services using thrift.
Currently, I am working as Graduate Research Assistant for Machine Intelligence Research lab at University of North Texas. I am exploring applications of Deep Learning in Natural Language Processing.

If your have a similar research interest or if you need any services related to one of these areas , please feel free to contact me.