This project is a backend implementation that automates the annotation function of VGG Image Annotator (VIA). It uses a trained neural network for inference, which significantly reduces the manual effort required in image annotation.
- Automated Annotation: The main feature of this project is to automate the process of image annotation. It uses a trained neural network model to predict annotations for a given image.
- Integration with VIA: This project is designed and tested to work with VGG Image Annotator (VIA).
- Specific domains: Currently only the heart slices dataset is supported shown in the demo.
- Client side: VIA run entirely on the clients browser. This makes deployment very easily manageable with serverless architecture. An inference server is needed for VIA-Auto to work.
docker pull tensorflow/tensorflow:latest-gpu-jupyter
docker run -it --rm -v $(realpath ~/notebooks):/tf/notebooks -p 8888:8888 tensorflow/tensorflow:latest-gpu-jupyter
To get started with this project, clone the repository and install the necessary dependencies.
Clone this repo
git clone https://github.com/bkutasi/VIA-Auto
Create an environment folder.
python3 -m venv env
Activate it
.\env\Scripts\activate
Install dependencies
pip install -r requirements.txt
Contributions are welcome! Please read our contributing guidelines before starting.
Since the original dataset seen in the demo was produced by SOTE researchers I wont release it. The documentation will contain everything you need to train your datasets besides bundled ones that are coming later.
- Full client side: VIA-Auto to run entirely on the client with WASM Tensorflow.js integration thus becoming part of the VIA package.
- VIA enhancements: VIA will be extended by some simple quality of life features that are deemed necessary.
This project is licensed under the terms of the GPLv3 license.