Machine Learning and Deep Learning Projects

Curated machine learning and deep learning projects spanning SVMs, Bayesian methods, ResNet experiments, and image captioning.

Overview

This repository contains selected projects from CG3201: Machine Learning and Deep Learning. The code has been cleaned into public-facing versions with reproducible entrypoints, while large datasets, checkpoints, caches, and assignment specification files are intentionally excluded.

Projects

• Project 1: SVM Image Classification Binary image classification using HOG features, a custom soft-margin linear SVM, and a kernel SVM comparison.
• Project 2: Bayesian and Generative ML Generative classification with Gaussian MLE / MAP and Laplace Naive Bayes, plus Bayesian linear regression benchmarking.
• Project 3: Tiny ImageNet ResNet Study Construction of a fixed 25-class Tiny ImageNet subset and comparison of four Mini-ResNet variants.
• Project 4: Image Captioning with Attention Flickr8k preprocessing, a ResNet-50 encoder, and a comparison between global-context and spatial-attention decoders.

Tech Stack

• Python
• PyTorch
• torchvision
• scikit-learn
• OpenCV
• NumPy
• pandas
• matplotlib

Repository Structure

machine-learning-deep-learning-projects/
  README.md
  project-1-svm-image-classification/
  project-2-bayesian-and-generative-ml/
  project-3-tiny-imagenet-resnet-study/
  project-4-image-captioning-with-attention/

Notes

• Dataset folders are not committed to this repository.
• Generated outputs, checkpoints, and temporary artifacts are excluded through .gitignore.
• Assignment specification PDFs and extracted prompt text are intentionally omitted.
• Each project folder includes a cleaned code version and the corresponding report PDF.