Support for rotating a rectangle feature

[patel-zeel]

Description

Motivation

Oriented bounding box (OBB) detection applications are increasing, especially in communities working on aerial/satellite imagery. Top computer vision conferences such as CVPR and ICCV have numerous papers proposing new methods to detect OBBs every year. OpenMMLab has a library mmrotate just for OBB detection. Few paperswithcode threads track the state-of-the-art in OBB detection [thread1, thread2].

Proposals

• P1: geemap drawing tools allow drawing a rectangle as well as a polygon. An OBB is something between the two. It needs 4 coordinates like a quadrilateral (Polygon) to represent itself correctly, but it has to be rectangular. Imagery labeling tools such as label-studio and Microsoft's satellite imagery labeling tool allow drawing a rectangle and then allow to rotate it. I think geemap has the necessary building blocks to make this happen and will immensely benefit the growing OBB community. Specifically, programmatic control of geemap makes it a very attractive tool for OBB labeling.

• P2: If geemap allows loading existing Polygons, not as a layer but an editable feature, it can be very useful to adjust wrongly predicted bounding boxes. Changing the class value is currently possible in leafmap via a solution proposed here.

[giswqs]

This is possible through the ipyleaflet Geoman Draw Control.
https://ipyleaflet.readthedocs.io/en/latest/controls/geoman_draw_control.html

import geemap
from ipyleaflet import Map, GeomanDrawControl

m = geemap.Map(center=(50, 354), zoom=5, draw_ctrl=False)

draw_control = GeomanDrawControl()
draw_control.polyline =  {
    "pathOptions": {
        "color": "#6bc2e5",
        "weight": 8,
        "opacity": 1.0
    }
}
draw_control.polygon = {
    "pathOptions": {
        "fillColor": "#6be5c3",
        "color": "#6be5c3",
        "fillOpacity": 1.0
    }
}
draw_control.circlemarker = {
    "pathOptions": {
        "fillColor": "#efed69",
        "color": "#efed69",
        "fillOpacity": 0.62
    }
}
draw_control.rectangle = {
    "pathOptions": {
        "fillColor": "#fca45d",
        "color": "#fca45d",
        "fillOpacity": 1.0
    }
}

m.add(draw_control)

m

[patel-zeel]

This is precisely what I thought with P1. Thank you, Prof. @giswqs! For P2, how to load previously saved features and edit their geometry?

[giswqs]

I am not sure it if supports loading existing vector data. You need to study their source code to see if it is feasible.
https://www.geoman.io/docs/modes/draw-mode#customize-style
https://github.com/jupyter-widgets/ipyleaflet/blob/master/python/ipyleaflet/ipyleaflet/leaflet.py#L2294

[patel-zeel]

Sure, Prof. @giswqs. I'll try to figure it out and if it works, will share here for everyone.

[patel-zeel]

Prof. @giswqs, how to access the rotated bounding box created by the Geoman Draw Control example shared above? I tried the following methods but they didn't work.

print(m.draw_control)
# None

print(m.draw_features)
# []