Personal

• K-Cryptocoin - An implementation of blockchain from scratch using the Node.JS packages able to create distributed network and all the feature for a blockchain required. And implemented cryptocurrency on top of this blockchain.

Graphics library development

• Rasterizing the polygon objects - Developed the graphics library for 2D and 3D transformations of the polygon images, Rasterizing the polygon objects, Clipping the polygons.
• Painters algorithm - Half-toning and gourmand shading on the polygon images.
• Raytracing - Implemented basic raytracer using OpenGL libraries.

Parallel computing and Big Data

• Gradient Descent and Logistic Regression - Developed Gradient Descent and Logistic Regression algorithms and optimize their gradient function so that it work with sparse date as well.
• Multiprocessing - Developed multiprocessing code which allows multiple process to take part in performing computation on the algorithm.
• Kmean - Developed K means clustering algorithm which allows sparse data matrix as input.

Computer Vision

• Content aware Resizing - Developed algorithm for image resizing by implementing Seam carving algorithm.
• Hough Transform - By making use of Hough transformation we can find the center of circle with fixed radius.
• Image Retrieval - Creation of visual vocabulary and retrieveing the images with similar patches as query image by using bag of words histogram.

Visual Recognition

• Human Ethinicity Prediction - CNN implemented using Keras with TensorFlow backend to identify ethnicity/race of a person . Achieved 84.5% accuracy with over 23,000 face images from the UTKFace dataset containing large variations in age, pose, expression, etc.

Work

Student intern, Stratovan Coorporation, Davis

• ComputerTomography - The implementation of the exsisting Randon Transformation method for CT Scan in C++. Which initially existed in Matlab. This project helps the company to compare their prototype which is implemented in C++ with the my code. This program utilizes Computer vision techniques for better CT scan images.