turktool
is a slick, modern React app for scalable bounding box annotation of images. You can connect it to Amazon Mechanical Turk and/or a custom backend to fulfill your deepest data labelling desires.
This project was bootstrapped with create-react-app
. You can find the original (and incredibly useful) README.md here.
Check out the live demo hosted on GitHub Pages.
- Redux undo and redo of bounding boxes using keyboard
- Crosshairs for maximum accuracy
- Toggling crosshairs using keyboard
- POSTing annotations to Mechanical Turk via a form
- GETting image from a URL
- Object classification is not support (I treat this as a separate task)