Kumar Siddhant

Kumar Siddhant

Graduate student at Rutgers University | Deep Learning | Machine Learning | ex-Developer @ Verizon Wireless”

Projects

Below are my projects!

RADICAL PILOT

September 05, 2018

Research Project, Rutgers University, Piscataway, NJ

In this project, I am researching and analyzing the efficiency of various scheduling algorithms for the Radical proprietary software namely Radical-Pilot. RADICAL-Pilot (RP) is a Pilot Job system written in Python. It allows a user to run large numbers of computational tasks (called ComputeUnits) concurrently on one or more remote ComputePilots that RADICAL-Pilot can start transparently on a multitude of different distributed resources, like HPC clusters and Clouds. Ater the analysis, I will be implementating the scheduling algorithms on Radical-Pilot.

CROWD IMAGES ANNOTATION AND ANALYSIS USING CASCADED CNNS

June 01, 2018

Research Project, Rutgers University, Piscataway, NJ

In this project, I collaborated with Prof. Vishal M. Patel and Mr. Vishwanath Sindagi to annotate approximately 4000 crowd images using the Amazon Mechanical Turk. Using the annotated images, we trained the cascaded CNN for Crowd Counting and Density Estimation.

PARALLEL COMPUTING

April 01, 2018

Project, Rutgers University, Piscataway, NJ

The objective of the project was to develop a parallel computing scenario using the server client architecture by implementing the socket programming algorithm.

THORACIC DISEASE PREDICTION USING CHEST X RAY IMAGES

April 01, 2018

Project, Rutgers University, San Francisco, California

In this project, a Convolutional Neural Network (CNN) is developed and trained to classify the thoracic diseases from the chest X-ray images that are available from the NIH database. The images available in the NIH database are labelled by Wang et al and are updated in a csv file. The objective of this project is to use the data and train a CNN from scratch or use a pre-trained CNN like AlexNet or VGG and classify the thoracic diseases from the chest radiographs.

FACE RECOGNITION

March 01, 2018

Project, Rutgers University, Piscataway, NJ

In this project, I Implementated the face recognition algorithm using AlexNet and VGG16 deep networks by training them from the LFW dataset. In the second part of the project, I implemented the face recognition algorithm using PCA, LDA, SVM SRC on YaleB dataset.

LOCATION PREDICTION USING COLLABORATED NETWORK IN SMARTPHONES

September 01, 2017

Project, Rutgers University, Piscataway, NJ

The algorithm that can perform prediction is based on a stochastic model on the HMM theory. This model is based on the concept of collaborative filtering where the algorithm predicts the context of the user based on the context observation of the users related to the primary user. This model can also use homomorphic encryption to preserve privacy of the users involved. In order to evaluate this model, we developed and used a context feature collection application to obtain data. Due to issues with one of the devices, only one device’s features were successfully collected. In order to evaluate the collaborative aspect of HCFContext, we synthetically create collaborators from this real dataset. Using this, we then evaluate the algorithm’s predictions against the sample data.

MAXIMUM POWER POINT TRACKING BY FRACTIONAL OPEN CIRCUIT VOLTAGE ALGORITHM USING A LOW COST MICROCONTROLLER

January 01, 2014

BTech Project, National Institute of Technology Rourkela, Rourkela, India

Solar energy extracted from a Photovoltaic cell offers an eco-friendly, free and renewable source of electricity which is still relatively costly and inefficient today which includes the difficulties related to complete harnessing of solar power. The maximum power point tracking (MPPT) of the PV panel for all sunshine conditions is the vital strategy to get the maximum power output thus increasing the efficiency of solar power extraction mechanism. The Fractional Open Circuit (FOC) Voltage Algorithm is implemented on a PV panel for MPPT by using a low cost microcontroller. In this paper the implementation of the FOC voltage method is shown. This method is simple and is used to approximate the linear relationship between the open circuit voltage and the MPP voltage for a PV panel. So, in this method, the current sensor is not required and the PV panel voltage is sensed using a voltage divider circuit. The effectiveness of the algorithm is verified with experimental results.

A STUDY ON THE ROBOT SWARM ANIMATION IN A SINGLE OBSTACLE WORKSPACE

May 01, 2013

Summer Internship Research Project, Indian Institute of Technology Guwahati, Guwahati, India

The Project is based on multi-robot based coordination and a study on their behavior as they come across an obstacle or any object in their path while moving towards a particular goal. The robots are programmed and are controlled by the intelligence programmed to them i.e. called the Swarm Intelligence. The simulations are first carried out in MATLAB considering a robot as a point. The points (the robots) are positioned at their respective positions and they have to reach the goal by bypassing an obstacle (a large sphere in this case). The trajectory obtained from the particles is saved for animation using Python-Ogre. The animation of the problem statement is done in Python-Ogre for proper visualization of the robots, their behavior and coordination. The robotic entities have to follow the trajectory obtained from the MATLAB simulation. The particles were intelligently controlled using Swarm Intelligence algorithms and were successfully simulated in MATLAB and animated in Python-Ogre.