🚀 Introduction

Welcome to the repository containing innovative software developed by Ultralytics 🧠. Our code is 🌟 open-sourced and freely available for redistribution under the AGPL-3.0 license. For more insight into our work and impact, head over to https://www.ultralytics.com.

📗 Description

The repository at https://github.com/ultralytics/mnist is our dedicated playground for the MNIST dataset. 🖐 This repository houses sandbox code that allows for experimentation and training of different neural network architectures on the famous MNIST digit database.

📦 Requirements

Ensure you have Python 3.7 or later installed on your machine. The following packages are required, and you can install them using pip with the provided command: pip3 install -U -r requirements.txt.

• numpy: A fundamental package for scientific computing in Python.
• scipy: Used by train.py to load the bundled MATLAB .mat MNIST files.
• torch: PyTorch, an open-source machine learning library for Python.
• torchvision: A PyTorch package that includes datasets and model architectures for computer vision.
• opencv-python: An open-source computer vision and machine learning software library.
• h5py: Provides HDF5 file support for reading large v7.3 .mat files in train_xview_classes.py.
• tqdm: Progress bars used by detect.py and train_resnet.py.

🏃‍♂️ Run

To start training on the MNIST digits dataset, install the dependencies and execute train.py from the repository root:

pip3 install -U -r requirements.txt
python3 train.py

The default script trains ConvNetb for 20 epochs using the bundled data/MNISTtrain.mat and data/MNISTtest.mat files, which were initially curated from Yann LeCun's MNIST database. Additional experimental entry points include train_resnet.py, train_sandd.py, and train_xview_classes.py.

💡 Contribute

We welcome contributions from the community! Whether you're fixing bugs, adding new features, or improving documentation, your input is invaluable. Take a look at our Contributing Guide to get started. Also, we'd love to hear about your experience with Ultralytics products. Please consider filling out our Survey. A huge 🙏 and thank you to all of our contributors!

📄 License

Ultralytics is excited to offer two different licensing options to meet your needs:

• AGPL-3.0 License: Perfect for students and hobbyists, this OSI-approved open-source license encourages collaborative learning and knowledge sharing. Please refer to the LICENSE file for detailed terms.
• Enterprise License: Ideal for commercial use, this license allows for the integration of Ultralytics software and AI models into commercial products without the open-source requirements of AGPL-3.0. For use cases that involve commercial applications, please contact us via Ultralytics Licensing.

📮 Contact

For bug reports, feature requests, and contributions, head to GitHub Issues. For questions and discussions about this project and other Ultralytics endeavors, join us on Discord!