PointRend is image segmentation as rendering
The PointRend neural network module performs point-based segementation predictions at adaptively selected locations based on an iterative subvision algorithm.
PointRend achieves higher sharpness on tricky object boundaries such as fingers than Mask R-CNN, and can be added on both semantic and instance segmentation.
This module show intermediate results. So if you want to use Point Rend, apply Point Rend to your instance segmentation or semantic segmentation project.
Ainize is done in the following steps:
- click 'default'.
- click 'try it out' and first, input mask-image file and second, input original-image file.
- click 'submit' button.
this is dockerized, it can be run using docker commands.
docker build -t pointrend .
docker run -d --rm -p 80:80 pointrend
Now the server is available at http://localhost:80.
