An open API service indexing awesome lists of open source software.

https://github.com/lmamon/vitransformer

GHW machine learning workshop using CIFAR10
https://github.com/lmamon/vitransformer

computer-vision machine-learning

Last synced: 6 months ago
JSON representation

GHW machine learning workshop using CIFAR10

Awesome Lists containing this project

README

          

# Vision Transformer (ViT) on CIFAR-10

deprecated and moved to HuggingFace https://huggingface.co/roylvzn/vit-cifar10
This repository contains the implementation of a Vision Transformer (ViT) trained on the CIFAR-10 dataset. This project was developed during the MLH Global Hack Week as an introduction to leveraging transformer models for computer vision tasks.

## Features

### Vision Transformer (ViT):

Implementation of a ViT from scratch.

Training and evaluation on the CIFAR-10 dataset.

### Deep Learning Frameworks:

PyTorch used for model creation, training, and evaluation.

## CIFAR-10 Dataset:

Used to benchmark the ViT's performance on image classification tasks.

## Reproducible Results:

Code for loading, training, and testing the model.

Pre-trained model checkpoint included for quick testing.

## Usage

1. **Clone the repository**:
```bash
git clone

3. **Navigate to the project directory**:
```bash
cd ViT-CIFAR10

3. **Run the Jupyter Notebook**:
```bash
jupyter notebook ViTransformer.ipynb

4. **For standalone scripts**:

Train the model: Use vit_model.py.

Load and test a pre-trained model: Use load_vit_model.py.

## Learning Objectives

Understand the architecture and implementation of Vision Transformers.

Learn to preprocess and train on CIFAR-10 using PyTorch.

Experiment with transformer-based models for image classification.

## Results

Accuracy: Achieved 93% accuracy with the CIFAR-10 test set.

Visualization: Includes attention maps and classification results.

## Contributing

Contributions are welcome! Please open an issue or submit a pull request with your improvements.

## License

This project is licensed under the MIT License. See the LICENSE file for details.