Exploring US Census microdata, tackling privacy issues, and anonymization. Exercise A delves into quasi-identifiers, anonymization methods, identification risks, and differential privacy. Exercise B involves data loading, k-anonymity, histograms, adding noise for privacy, computing private averages, and analyzing privacy parameter impacts.

Image Processing Course - HW3

Develop a simulation platform1 for a BPSK, 4QAM, 8PSK and 16QAM communication system transmitting information over an additive white Gaussian noise (AWGN) channel.

Program in Matlab R2021b that makes noise to different images.

Explored Machine Learning regression models of varying flexibility and how flexibility relates to MSE, Bias, and Variance in our predictions for MATH 4377: Math of Machine Learning. Visualizations of the Bias-Variance trade-off are included, and the project heavily relied on Spline Regression degrees of freedom for flexibility measuring.

GPU-ported Non-Local-Means for accelerated image denoising

The standard approach to image reconstruction using deep learning is to use clean image priors for training purposes. In this project, we attempt to achieve denoising without using a clean image prior and yet, achieving a performance comparable to, or sometimes, even better than that obtained using the conventional approach.

A GUI and API for Gaussian and pseudo noise addition, and filtering with vector-based filters: Basic Vector Directional Filter, Vector Median Filter, Distance Directional Filter.

Python implementation of the Non Local Means algorithm for Image Denoising

Specifically showed that eigenvalue variations under a perturbation are not a good measure of the overall sensor performance near an Exceptional Point.

I built a Denoising Autoencoder to remove noise from the image. Image Denoising is the process of removing noise from the Images The noise present in the images may be caused by various intrinsic or extrinsic conditions which are practically hard to deal with. The problem of Image Denoising is a very fundamental challenge in the domain of Image …

Developed robust image classification models to prevent the effect of adversarial attacks

Depth peeling for point clouds.

Official implementation for "Pure Noise to the Rescue of Insufficient Data: Improving Imbalanced Classification by Training on Random Noise Images" https://arxiv.org/abs/2112.08810

Course projects, capstone and individual studies.

Video Denoising using Low Rank Matrix completion

Detection and removal of specific types of outliers present in different data formats which includes detection and removal of contextual outliers from textual data using LOF, outliers from tabular numeric data using LOF, gaussian noise from image data using NLM.

In this project, low-pass filters and Kalman filters with different window function designs are used to denoise speech signals polluted in the full frequency band of Gaussian white noise

Add Gaussian noise to point cloud