Improved GUI for training sample collection

[giswqs]

The current implementation can be improved by creating all classes beforehand and users can select whatever class to create training samples.

62_training.mp4

[iamrajbhattarai]

Hello Professor,

I was trying of perform supervised classification inside Nepal.

Map = geemap.Map()

ROI = ee.Geometry.Polygon(
  [[[85.27443103361846,27.629026415947422],
    [85.41107348967314,27.629026415947422],
    [85.41107348967314,27.737866221318047],
    [85.27443103361846,27.737866221318047],
    [85.27443103361846,27.629026415947422]]])

image = ee.ImageCollection('LANDSAT/LC08/C01/T1_TOA') \
    .filterBounds(ROI) \
    .filterDate('2020-01-01', '2021-01-31') \
    .sort('CLOUD_COVER') \
    .first() \
    .select('B[1-7]') \
    .clip(ROI)  


vis_params = {
    'bands': ['B5', 'B4', 'B3']
}

Map.centerObject(ROI, 12)
Map.addLayer(image, vis_params, "Landsat-8")

Map

Then I started collecting samples as done in this video: https://www.youtube.com/watch?v=VWh5PxXPZw0
However, those features don't get stored.

But, when I draw casually (when not in collecting sample mode), it stores the object.
Please, do help me.

[giswqs]

@iamrajbhattarai You can use Map.user_rois to return the training samples as a feature collection.

[iamrajbhattarai]

Yes, I did that. But, the objects are not returned as feature collection. I think there is problem in the version available on the website or git hub. Even the polygons are joint when we draw. The polygon drawn during interaction in sample collector does not seem to work as shown in your video. After we draw a polygon, next polygon also starts from the end point of last polygon. But, main problem is those features are not returned as feature collection when using Map.user_rois.getInfo() command.

[giswqs]

I encountered this weird issue before. Try downgrading ipyleaflet to an earlier version, e.g., pip install -U ipyleaflet==0.13.3

[iamrajbhattarai]

Thank you professor. I made it.