portfolio

what I do as an engineer

  • publications
  • projects
  • industry

2017

  • Analyzing ConvNets Depth for Deep Face Recognition

    Mohan Raj, I. Gogul, M. Deepan Raj, V. Sathiesh Kumar, V. Vaidehi, S. Sibi Chakkaravarthy

    CVIP-2017, Springer pp 317-330

    [paper]
  • Flower species recognition system using convolution neural networks and transfer learning

    I.Gogul, V. Sathiesh Kumar

    ICSCN-2017, IEEE Xplore

    [paper]
  • Static gesture recognition based precise positioning of 5-DOF robotic arm using FPGA

    M. Deepan Raj, I. Gogul, M. Thangaraja, V. Sathiesh Kumar

    TIMA-2017, IEEE Xplore

    [paper]

2016

  • Smart Autonomous Gardening Rover with Plant Recognition using Neural Networks

    V. Sathiesh Kumar, I.Gogul, M. Deepan Raj, S.K.Pragadesh, J. Sarathkumar Sebastin

    ICACC-2016, Elsevier Procedia Computer Science Volume 93, 2016, Pages 975-981

    [paper]


2018

  • DeepDrum & DeepArp

    Used Google Magenta's DrumsRNN and ImprovRNN to generate drum patterns and arpeggio patterns based on user's seed pattern. Created timeline and multiple pattern generation in a single browser window using JavaScript.

    Tools used: HTML5, CSS3, JavaScript, Magenta.js, TensorFlow.js, Tonal.js, jquery.

    demo | video | code | tutorial
  • Emotion Recognizer using Deep Neural Network

    A real-time implementation of emotion recognition using two deep neural networks (extractor and classifier) using Google's TensorFlow.js in the browser. Model is created, trained and inferred in real-time with data acquisition happening in client's device.

    Tools used: TensorFlow.js, HTML5, CSS3, JavaScript, jQuery, Sass.

    demo | video
  • Recognizing Digits using Deep Neural Network in Google Chrome

    Recognize handwritten digits drawn by a user in a canvas in real-time using Deep Neural Network such as Multi-Layer Perceptron (MLP) or Convolutional Neural Network (CNN) in the browser (specifically Google Chrome).

    Tools used: Keras, TensorFlow.js, HTML5, CSS3, JavaScript, jQuery, Sass, Python.

    Dataset: MNIST Handwritten Digits

    video | tutorial
  • Classifying images using Keras MobileNet in Google Chrome

    Perform image classification in real-time using Keras MobileNet, deploy it in Google Chrome using TensorFlow.js and use it to make live predictions in the browser (specifically Google Chrome).

    Tools used: Keras, TensorFlow.js, HTML5, CSS3, JavaScript, jQuery, Sass, Python.

    Dataset: IMAGENET (1000 categories)

    tutorial

2017

  • Flower Species Recognition System

    Recognize different flower species using state-of-the-art Deep Neural Networks such as VGG16, VGG19, ResNet50, Inception-V3, Xception, MobileNet in Keras and Python. Also, a detailed comparison between Global Feature Descriptors and data-driven approach for this fine-grained classification problem was studied.

    Tools used: Keras, Python.

    Dataset: FLOWERS17 (University of Oxford)

    video | tutorial 1 | tutorial 2
  • Sound Classification using Neural Networks

    An environment sound classification example that shows how Deep Learning could be applied for audio samples.

    Tools used: Keras, Python.

    Dataset: ESC-50 - Environmental Sound Classification

    video

2016

  • Monocular Visual Odometry using OpenCV and Python

    Feature based Monocular Visual Odometry using FAST corner detector, KLT Tracker, Nister's five point algorithm and RANSAC algorithm with the help of OpenCV and Python.

    Tools used: OpenCV, Python.

    Dataset: KITTI

    video
  • SMART Home - Continuous Voice Recognition using Android

    A SMART Home automation system using off-the-shelf technologies such as Android and Arduino to control home appliances such as Fan, Light bulbs and other electronic appliances with the help of relay and your voice.

    Tools used: Arduino Uno micro-controller, Android smartphone, 8-channel relay module, HC-05 Bluetooth module, Jumper wires, Batteries, Arduino IDE, Android Studio 2.2, Philips Wireless speaker.

    video
  • SLAMINOR - Parallel DC and Servo Control using Xilinx Zedboard

    Parallel control of 2 DC motors and a servo motor using Xilinx Zedboard.

    Tools used: FPGA - Xilinx Zedboard, IDE - Vivado Design Suite 2014.2, Clock Frequency - 50 MHz, DC motors - 500 RPM 12V, Servo motor - Futaba S3003, Battery - 12V 1.3A, Motor Driver - L293D.

    video
  • Hand Gesture Recognition and Servo Control

    Recognize hand gestures using OpenCV and Python, and control a servo motor based on the gestures using Odroid-XU4 and Arduino Mega.

    Tools used: Ardunio Mega, Odroid-XU4, Python, Arduino IDE, Servo motor - Futaba S3003, Battery - 12V 1.3A.

    video | tutorial 1 | tutorial 2
  • Medical Quadcopter

    A standard Quadcopter for medical applications.

    Tools used: Flight Controller - APM 2.6, Electronic Speed Controllers - 30A, Brushless DC Motors - 1000KV, Power Source - Turnigy 3000 mAh 3S 20C LiPo battery, Quad Copter Frame - F450, Turnigy 6 channel FHSS 2.4Ghz Tx/Rx.

    video
  • Autonomous LOTA Robot

    A small robotic vehicle that can follow a line, detect obstacles, manages to run on the top of a table without falling down and could control its speed with the help of sensors and ADC.

    Tools used: Microcontroller - ATmega16, DC Motors - 100 RPM, Power source - 12V battery, Sensors - 4 Infrared sensors, Other parts - Potentiometer, NOT gate, chassis, wheels.

    video


Qualcomm

Designation Engineer
Location Chennai
Period Jun 19, 2017 - Present
Primary Domain Physical Design & Signoff
Other Domains Automation, Machine Learning, Deep Learning, Web Development, Tool Development
Languages Python, Tcl, Shell, Perl, HTML5, CSS3, JavaScript

Nokia

Designation Graduate Technical Intern
Location Chennai
Period Dec 05, 2016 - Jun 05, 2017
Primary Domain Computer Networking
Other Domains Automation, Tool Development, Machine Learning, Web Development
Languages Python, Tcl, Shell, HTML5, CSS3, JavaScript

TATA Consultancy Services

Designation Assistant Systems Engineer
Location Chennai
Period Jun 26, 2014 - Aug 15, 2015
Primary Domain Front-End Web Development
Other Domains Web Development, Android App Development, Hybrid Mobile App Development, User Interface Design
Languages HTML5, CSS3, JavaScript, jQuery, Java, Android, .net, C#