This is a tool for object labelling, that takes special features of satellite images into account. It allows tile-by-tile image browsing, different band combinations veiwing and 'magic-wand'-like object selection.

• Clone repository
• Set sat-img-lab as a current directory
• Create virtual environment
• Source virtual environment
• Install requirements
• Run script

$ git clone https://github.com/WGussev/sat-img-lab.git
$ cd sat-img-lab
$ python3 -m pyenv envname
$ source envname/bin/activate
$ pip install -r < requirements.txt
$ python3 labeller.py

1. Open a folder containing band images (developed with Sentinel-2 spectral channels in mind).
2. Open a folder with mask (.npy) files.
3. Choose or create a mask to work with. (you should have a separate mask for every object type)
4. Browse images with previous/next buttons.
5. Choose different image band combinations with the combobox.
6. Choose objects and add them to the mask.

The tool is under developement. Feedback appreciated.