CIFAR-10-ResNet-Global-Optima

Welcome to the Mini-project repository for the Global Optima group. This project focuses on optimizing and modifying the ResNet architecture to achieve the highest possible accuracy on the CIFAR-10 image classification dataset while maintaining a model size of no more than 5 million parameters.

Project Deliverables

results/

This directory contains all the key components of our project:

• Report.pdf: Our comprehensive project report detailing the methodologies, experiment results, and conclusions.
• resnet_5M.ipynb: The Jupyter notebook for our final ResNet model with 5 million parameters.
• predictions.csv: The predictions file that was used for the Kaggle competition submission.
• ckpt.pth: The checkpoint file containing the trained model weights.
• training_validation_plot.png: Visual representation of training and validation accuracy and loss over the epochs.

Experiments/

Contains Jupyter notebooks for various configurations and experimental setups tested during the development phase of the project.

Getting Started

To get started with exploring and running the project:

git clone https://github.com/tanmayr71/CIFAR-10-ResNet-Global-Optima.git
cd CIFAR-10-ResNet-Global-Optima