Improve image segmentation interface

[deneland]

Currently, image segmentation still feels a bit 'clunky' and is only usable for the most basic tasks. Some ideas to improve the experience:

• Images are presented in a relatively small container, especially when working with lower resolution images. In these cases both zooming and increasing the image size in the labelling setup cause the image to be 'cropped', while the majority of the screen real estate remains unused. Below an example with from top to bottom: 'vanilla', zoomed, and width set to 150%.
• Panning with the middle mouse button and a label selected generates mouse clicks, thus new regions.
• Being able to combine different tools such as polygons and brushes together is extremely valuable, by for example allowing you to first 'rough out' a shape using the polygon tool and then refining with brushes & erasers
• Content-aware tools such as a magic wand would greatly speed up segmentation.

[mobiuscreek]

Thank you. This would be a good addition.

[BradNeuberg]

+1 on magic wand support, for our use case that would dramatically improve semantic segmentation labeling.