Build your own image classification service within a day.
Zita is an all-in-one machine learning solution for building production-ready image classification services. It covers the full life cycle of creating a simple image classifier just like Amazon SageMaker.
- An easy-to-use image labelling tool for creating ground truth
- Out-of-the-box Jupyter notebooks and CLI commands to efficiently train and tune modern deep learning image classifiers.
- High-performance RPC and web API to test, evaluate and serve predictions.
With only minimal setup, anyone can build, train, and deploy a production-ready custom image classification service within hours.
We recommend install from source code:
git clone http://github.com/ktmud/zita.git
# Install Python packages
cd python
pip3 install poetry
poetry install
# Enter the Python envrionment managed by Poetry
poetry shell
# Go back to Zita root
cd ../
npm install
npm run devThis will start the tagging service first. You'd need to tag the photos in the Label tab, train models in a notebook, then be able to come back and view training predictions in the "Test" tab.
For details, check out SETUP.md.
MIT