Shiva Sanketh R M

Worked on algorithms - A*(A star), Particle Swarm Optimization (PSO) and Rapidly Growing Random Tree (RRT) to achieve Path and Motion planning. These two algorithms combine the essentials of an exploration algorithm with that of optimizing one to deliver results that are dynamic and adaptable.

Role - As a fresher I implemented the algorithms, also providing assistance in optimization and carried out simulation tasks in the project

• Client: Tata Elxsi
• Date: Feb - July 2017
• Type: R & D

Neural Artistic Style Transfer is a technique for stylizing content images with styles from a separate style image. Fast Neural Style Transfer is a modified technique in which a feedforward network learns stlizations for a style image beforehand so that it can stylize images instantly.

Role - Using the existing VGG16 model as a base model, implemented style transfer using VGG19, producing a comparable analysis between the two models for Style Transfer. Also implemented the algorithm to work on videos. Introduced another layer called the texture layer for the existing two layer model. This texture layer resulted in a more artistic look compared to the base model.

Libraries/Platform - Machine Learning, Deep Learning, Computer Vision, Keras, Tensorflow, CNN

• Type: Open Source
• Date: April 2019

Project - Chatbot - "Alex"

Developed an interactive chatbot embedded robot that can provide visual response for queries sent over Bluetooth. The robot is capable of movement based on the interpretation of received messages using the chatbot. "Alex" was used in the college technical fest to smartly answer queries of the attendees

Libraries/Platform - Python, Natural Language Processing, NLTK, Regular Expressions,

• Type: Undergraduate Project
• Date: October 2015

Project - Simulaneous Localization and Mapping based on Deep Neural Networks for Point Cloud Data

On-going Research Project

Libraries/Platform - Open3d, Pytorch, Python

• Type: Research (On-going)
• Date: April 2019 -

Project - Autonomous Vehicle Development Project

Achieved 84% accuracy on Object Detection and Tracking for Pedestrian Tracking and Vehicle Tracking using SSD at 22 FPS on NVIDIA 1080Ti GPU.

Libraries/Platform - Python, Keras, OpenCV, Deep Learning, Computer Vision, Linux

• Client: TATA ELXSI LTD
• Date: September 2018

Project - Autonomous Vehicle Development Project

Assisted in the development of Adaptive Cruise Control and Stop-and-Go feature based on Lane Detection and speed of the vehicle ahead.

Also worked on integrating different sensor modules on ROS for data synchronization

Libraries/Platform - ROS, Python, OpenCV, Tensorflow

• Client: TATA ELXSI
• Date: Nov 2017 - March 2018

Project - Driver Monitoring System

Role - Developed the following for the pipeline - Head pose Estimation, Eye gaze detection and Tracking, Eye Open percentage Estimation, Facial Emotion Recognition and Driver Drowsiness Detection achieving upto 20 FPS on NVIDIA 1080Ti with an input of 30 FPS.

Libraries/Platform - Python, OpenCV, Keras, Machine Learning, Deep Learning, Time Series Analysis

• Client: TATA ELXSI
• Date: June - December 2018

Project - CanSat Development(Miniature Satellite that fits in a can)

Worked on processing of images captured by the CanSat. Also worked on different sensor modules on Arduino

• Client: OpenCube Labs (Start-up)
• Date: Jan 2015 - May 2015

Project - Laser Synth (Music Synthesizer)

Designed and built an Electronic music synthesizer consisting of a system of Lasers and a Microcontroller. Obtained Provisional Indian patent for the same (Indian patent number: 5612/CHE/2014)

• Type: Undergaduate Project
• Date: October 2013

Project - Data Analysis and Sentiment Analysis

Implemented a project on analyzing large-scale data and deriving insights from Airbnb dataset of New York City. Developed rule-based Sentiment Analysis to derive user insights from user review data

Libraries/Platform - Machine Learning, Data Science, Python, matplotlib, NLTK, Data Analysis.

• Type: Open Source Project
• Date: August 2019

Project - Text Prediction and Poem composer

Developed Text Prediction algorithm without making use of any NLP specific libraries, capable of composing appreciably meaningful text paragraphs. Modified the same text predictor which allows users to compose Shakespeare like poem with the algorithm

Libraries/Platforms - Natural Language Processing, Python, PyTorch, Data Science, RNN, LSTM

• Type: Open Source Project
• Date: September 2019

Project - Brain Computer Interface

Developed four algorithms for Brain State Estimation, Attention Level Estimation, Eye Blink Detection and Jaw movement Detection. Also devised a mind controlled bot that was activated on a person's attention level.

• Type: Undergaduate Project
• Date: October 2014 - March - 2016

Project - Web Development

Developed a Technical Profile Website

Tools - HTML5, CSS3, JavaScript

• Type: Open Source Project
• Date: August 2019

Project - Audio Synchronized LED display

Developed an audio Synchronized LED display that changes color based on the music pattern being played. Uses Signal processing agorithms

• Type: Undergaduate Project
• Date: March 2014