GraceSevillano / Advanced-Image-Analysis-Assignments

Solutions for Advanced Image Analysis course assignments, featuring model designs for image summation and generation with MNIST, and style transfer using CycleGAN with MNIST and SVHN datasets.

Advanced Image Analysis Assignments

This repository contains my solutions for two assignments from the Advanced Image Analysis course, taught by Prof. Renato Martins. The assignments involve complex tasks in image processing, leveraging neural networks and generative models.

Assignments Overview

Assignment I - Learning to Sum and Generate Images of Numbers

• Objective: Design models (MLPs, CNNs, autoencoders) to understand, add, and generate images of numbers using the MNIST dataset.

• Contents: Jupyter notebook with step-by-step solution, datasets, output images, trained models, and a detailed report.

Assignment II - Style Transfer with Unpaired Collections of Images for Classification

• Objective: Perform classification and style transfer using unpaired image samples from the MNIST and SVHN datasets, inspired by CycleGAN.

• Contents: Jupyter notebook with detailed implementation, datasets, style-transferred image outputs, trained models, and an extensive report explaining the solution.

Getting Started

1. Clone this repository.
2. To set up the environment for each assignment, navigate to the respective assignment directory and install the required Python packages using the appropriate requirements file:
   • For Assignment 1: pip install -r Assignment_1/requirements_1.txt
   • For Assignment 2: pip install -r Assignment_2/requirements_2.txt
3. Explore the Assignment_1 and Assignment_2 directories for notebooks, datasets, and reports.

Technologies

• Python
• PyTorch
• CycleGAN for style transfer
• MNIST and SVHN datasets

Acknowledgements

Special thanks to Prof. Renato Martins for guiding these intricate projects, and to the community for the resources and tutorials on PyTorch and neural network training.