Projects

Sign Language Translator from Video to Speech - 2019

The goal of this project was to recognize the sign Language from video and generate simple English sentences.

Created the sign language data set for 30 sign words and 24 alphabets which could be classified using just one image frame.
Developed a deep learning model for classifying sign language images to corresponding sign words by transfer learning of the Inception V3 architecture and extended it to videos for generating simple English sentences
Got an accuracy of 62.5% on the test set under normal lighting conditions.

Customized Adversarial Image Generator - 2018

The goal of this project was to implement fast and efficient Customized Adversarial Image Generator which can generate adversarial images for MNIST dataset.

The User has the option to specify the input image and the target class, the perturbed input image which will be classified as target class will be produced.
Implemented a variation of Fast Gradient Sign Method (FGSM) algorithm to perturb the input image.
The Perturbed image which will be indistinguishable to input image by Human eye are produced.
Reference : Explaining and Harnessing Adversarial Examples by Ian J. Goodfellow et al.

View Project

Credit Card Fraud Detection - 2018

The goal of this project was to predict whether a given credit card transaction is fradulent or not using anomaly detection algorithm.

Used multivariate gaussian distribution to model the probability density function, which is then used to flag whether a transaction is fradulent or not.
Used Feature engineering to remove the features that didn't have much impact in modeling the density function.
Dataset source: Kaggle (https://www.kaggle.com/mlg-ulb/creditcardfraud)

View Project

ROOT-TMVA (Toolkit for Multivariate Analysis) - 2018

The goal of this project is to implement the new deep learning optimization algorithms in the ROOT-TMVA Project.

Refactored the existing code to include support for new deep learning optimization algorithms.
Implementing the optimization algorithms like SGD, RMSProp, Adam etc.
Unit tests and Integration tests are written to check the optimization algorithms.

View Project

Docdroid - 2017

The goal of this project was to develop an android application that can help people in medical emergency situations like a heart attack or an accident.

Created an Android application that can be used by anyone during potential medical emergencies or accidents.
With a single tap, an ambulance will be automatically booked and the nearest hospital having adequate facility will be allocated to the patient.
Patient’s emergency contacts will receive notification about him and they can live track him, the status of ambulance etc.
The server side of the application is build using Flask framework.

View Project

Online Assessment System for College Students - 2017

The goal of this project was to develop an Online Assessment System for college students.

Created an web application using Flask and Bootstrap framework.
Faculty can create their own tests and students can take them online.
It automatically evaluates both objective and descriptive answers and send the results to both student and faculty.
Faculty can view the list of students who have taken the test along with the marks obtained.

View Project

Alpha Math Tools - 2015

The goal of this project was to develop an Android application for doing basic math.

The app was created using Java and XML.
It provides facilities for calculating interests for automobile/home loans.
It also provides facilities for school students to solve linear, quadratic equations, etc.
It is available on Play Store as “Alpha Math Tools” (Rating : 4.9 , Downloads: 100+ )

View Project