List of major projects

This page contains a list of major projects I have done after 2021, most of which have been in Computer Vision, Machine Learning, Communication topics, etc. Each entry has a brief description of the project like the tech stack used and so on. The links to individual project pages will be included later; The list will grow with time!

  1. Information Flow Across Neural Network Layers

This project aims to decode the black box nature of neural networks and provide insights on how information is processed over the layers of a network based on many metrics such as information capacity. Applications include and are not limited to ease of transfer learning, reduction in training time and domain adaptation.

  1. Retinal fundus image enhancement with image decomposition

Image decomposition conducted to divide an input image into three layers: base, detail and noise layers. Illumination correction, detail enhancement and denoising are conducted respectively at these three layers. Follows this paper.

  1. Implementation of background subtraction on video from scratch

Based on “Adaptive background mixture models for real time tracking” by Stauffer and Grimson. Each pixel is modelled as a Mixture of Gaussians (MOG) to segment foreground.

  1. Comparison of Eigen Faces and Fisher Faces algorithms

Facial recognition algorithms comparison (unsupervised vs supervised) with 40 classes from AT&T dataset. Fisher faces fared better with zero error.

  1. Polynomial Curve Fitting and Genomic Sequence Analysis

Polynomial curve fitting is a regression problem implemented from scratch to best fit the data. Hyper-parameters tuning was the main focus. Genomic Sequence Analysis is a binary classification problem, solved with LibSVM library. Experimented with linear and gaussian kernels. Accuracy achieved was 95%.

  1. Image Processing Techniques for Chest Radiography Enhancements and Pneumonia Detection

This project attempts to create an expert classifier system for pneumonia diagnosis. Uses Keras framework and deployed with Flask back-end. Trained on 5000 chest X-ray images and achieves a 97% train accuracy and test accuracy of 93%.

  1. LZ77 Encoding-Decoding

Python implementation of the classic Lampel Ziv 77 encoding and decoding algorithm on randomised English text.

  1. Loyd Max Entropy Quantisation

Python implementation of Lloyd Max algorithm for entropy based quantisation of a random source following three distributions: Uniform, Gaussian and Rayleigh.

  1. FRAS: Facial Recognition Attendance System

Project was developed on DLIB, a facial recognition library in Python and successfully deployed in a class for attendance, with both android and website versions. Has capability to register 100 students per class without errors.

  1. Remote Server Health Monitoring System

Python based website app that pings servers at intervals and reports anomalies. Was used by NALCO as a client.