Selected Academic Projects


Predicting COVID-19 From Chest X-Ray Images

  • Built a deep learning pipeline to predict COVID-19 from chest X-ray images using transfer learning on pre-trained SqueezeNet.
  • Applied data augmentation and preprocessing to improve generalization on limited medical data.

Tile Crack Detection with U-Net

  • Developed an image segmentation model to detect surface cracks on tiles using U-Net.
  • Combined classical image processing with deep learning to localize fine-grained defects.

Genetic and Ant Colony Algorithms

  • Implemented a Genetic Algorithm to solve cubic equations and explore convergence behavior.
  • Implemented an Ant Colony Optimization algorithm for task scheduling problems.

Football Match Fuzzy Predictor

  • Designed a fuzzy logic controller to predict football match outcomes using interpretable rules.
  • Used the Simpful library to define linguistic variables and construct fuzzy rule bases.

Image Recalling Hopfield Network

  • Implemented a Hopfield network for recalling patterns on Fashion-MNIST–style image data.
  • Extended the Hopfield formulation to handle a TSP-like optimization setting.

SOM and RBF Clustering

  • Implemented Self-Organizing Maps (SOM) and Radial Basis Function (RBF) networks for classification tasks.
  • Used TensorFlow and Keras to train and compare different clustering and representation schemes.

MLP From Scratch

  • Implemented a multilayer perceptron from scratch for classification problems.
  • Used Keras to experiment with momentum and weight decay for optimization and regularization.

Artificial Intelligence Course Projects

  • Built a deep-learning–based model to recognize different clothing types from images.
  • Completed UC Berkeley–style Pac-Man AI projects, covering search, reinforcement learning, CSP, and MDP formulations.

Smart Sensor-Embedded Games for Autism and Alzheimer’s

  • Co-designed smart games to support Alzheimer’s patients and autistic children, focusing on engagement and cognitive support.
  • Developed sensor-embedded puzzles and interactive game rooms to assist caregivers with monitoring and assessment.
Niki Nezakati
Niki Nezakati
CS PhD Student