The Jakarnotator is an annotation tool to create your own database for instance segmentation problem. Jakarnotator data are compatible with the format of MS Coco database.
The Jakarnotator is based on Leaflet. You can annotate any image format supported by Leaflet (.png, .jpeg, ...) and even georeferenced data. You just need to put those images into a specific folder. Getting started
At Jakarto, we used a Jakarnotator instance to create a database of urban furnitures. With these data we were able to train our models with the Mask RCNN algorithm (no modification needed to deal with input data format as Detectron (a Mask RCNN implementation) used MS Coco).
A live demo is available on jakarnotator.jakarto.com.
To see the Jakarnotator in action please take a look to these two links :
Click here if you want to learn how to annotate objects in the Jakarnotator
If you are even more interested in creating your database from your own images or create new classes of objects you should take a look to this GETTING_STARTED.md.
Jakarnotator is open-sourced as a simple tool to ease the creation of instance segmentation databases. Feel free to open an issue for questions, details, new features...