Tools in this repository are designed to allow the user to train his own Mask R-CNN model in GRASS GIS and to apply it to get vector masks from raster data. This repository is a follow-up development of a project created for my master's thesis (see here). The library for these tools is heavily based on Python implementation of Mask R-CNN by Waleed Abdulla, Matterport, Inc. (see here).
In GRASS GIS, the stable version of i.ann.maskrcnn.* modules can be installed
from the official AddOns repository by command:
g.extension extension=i.ann.maskrcnn
To install the working version from this repository, clone it and install into
GRASS with command:
g.extension extension=i.ann.maskrcnn url=path/to/the/i.ann.maskrcnn/folder
The official documentation is on the official web site.