Nishai Kooverjee

Masters student at Wits University, Johannesburg.
Helping machines learn stuff.

Yo!โœŒ๐Ÿฝ

Hi. I'm Nishai. I am currently doing my Masters in Computer Science at Wits University, where I am a member of the RAIL research lab. I find it fascinating how deep learning algorithms are able to learn really complex things. I'm interested in disentangling their structure and understanding how they manage to achieve such impressive feats.
I live in Johannesburg, South Africa. I consider myself focused, ambitious and creative. Besides machine learning, I love listening to music, reading, travelling, meeting new people and discovering new ideas.

Education ๐Ÿ‘จ๐Ÿพโ€๐ŸŽ“

2020 -

MSc Computer Science


University of the Witwatersrand, Johannesburg

2019

BSc Honours, Computer Science (cum laude)


University of the Witwatersrand, Johannesburg

2016 - 2018

BSc Computer Science (cum laude)


University of the Witwatersrand, Johannesburg

Publications ๐Ÿ’ก

Inter- and Intra-domain Knowledge Transfer for Related Tasks in Deep Character Recognition ยท

arXiv

ยท

IEEE Explore
Published in PRASA 2020
Show Abstract

Pre-training a deep neural network on the ImageNet dataset is a common practice for training deep learning models, and generally yields improved performance and faster training times. The technique of pre-training on one task and then retraining on a new one is called transfer learning. In this paper we analyse the effectiveness of using deep transfer learning for character recognition tasks. We perform three sets of experiments with varying levels of similarity between source and target tasks to investigate the behaviour of different types of knowledge transfer. We transfer both parameters and features and analyse their behaviour. Our results demonstrate that no significant advantage is gained by using a transfer learning approach over a traditional machine learning approach for our character recognition tasks. This suggests that using transfer learning does not necessarily presuppose a better performing model in all cases.

Some Projects ๐Ÿ–ฅ

Snake classification challenge using Convolutional Neural Networks

GitHub

This involved classifying images of snakes into their respective species, as per the AICrowd challenge. Various pretrained architectures were used, including AlexNet, MobileNet-v2 and ResNet-18. This project was done for my Computer Vision course.


Reinforcement Learning agent for Obstacle Tower

GitHub

We trained a reinforcement learning agent using a Double DQN (Deep Q Network) to play the Obstacle Tower game. This was a modified version of the Obstacle Tower challenge, done in a group for my Reinforcement Learning course.


Person detection using SVMs

GitHub

A system to detect people in images or a live video stream. I personally collected images of each class type (person and not-person), and trained linear Suport Vector Machines (SVMs) using Histograms of Oriented Gradients (HOG) as feature descriptors. Includes a live demo using webcam.


...Check out more stuff on my GitHub! โœŒ๐Ÿฝ

Skills & things ๐Ÿ‘จ๐Ÿพโ€๐Ÿ’ป

Basic Good Proficient

Languages

Python


Machine Learning stuff
  • PyTorch
  • PyTorch-Geometric
  • Tensorflow
  • Sci-kit-learn
  • Dev
  • Django
  • Wagtail
  • Other libraries
  • Pandas
  • Matplotlib
  • Seaborn
  • NetworkX
  • Java


    IDEs and Frameworks
  • Android Studio / IntelliJ Idea
  • Eclipse
  • C++


    Parallel computation
  • OpenMP
  • HTML / Javascript


    JS libraries
  • JQuery
  • Bootstrap 4
  • Frameworks
  • Vue.js
  • React.js
  • Graphics
  • Three.js
  • C


    Parallel computation
  • OpenMP
  • Graphics
  • OpenGL
  • Matlab


    C#


    Frameworks
  • ASP.net
  • Xamarin.iOS
  • Xamarin.Android
  • Swift


    Experience ๐Ÿ’ผ

    February 2020 -

    Computer Science Head Tutor


    University of the Witwatersrand, Johannesburg

    September 2019 -

    Freelance Web Developer


    February 2019 - November 2019

    Computer Science Tutor


    University of the Witwatersrand, Johannesburg

    July 2018

    Intern developer


    Retro Rabbit, Johannesburg

    June 2017

    Intern developer


    BBD, Johannesburg

    Get in touch

    Reach out if you have something interesting to chat about or just want to grab coffee! โ˜•๏ธ