In [ ]:

    

# Load code to setup authorization for an IPython Notebook. Note that this is a temporary step
# that is required until the Earth Engine Python API is updated to include this logic.
%run 'authorize_earth_engine_in_notebook.ipynb'

# Initialize Earth Engine
# Note that we are calling a function defined in the previously run IPython Notebook, rather than
# the typical call to ee.Initialize()
ee_initialize()


    

In [ ]:

    

%run 'define_google_maps_interactive_widget.ipynb'


    

In [ ]:

    

map = GoogleMapsWidget(lat=31.073, lng=30.732, zoom=8)
display(map)

composite = ee.Image("L7_TOA_1YEAR_2000")

bands = ["10", "20", "30", "40", "50", "62", "70"]

# Use the MCD12 land-cover as training data.
cover = ee.Image("MCD12Q1/MCD12Q1_005_2001_01_01").select("Land_Cover_Type_1")

# A region of the image to train with.
region = ee.Geometry.Rectangle(30, 30, 32.5, 31.6)

PALETTE = ','.join([
    'aec3d4', # water
    '152106', '225129', '369b47', '30eb5b', '387242', # forest
    '6a2325', 'c3aa69', 'b76031', 'd9903d', '91af40', # shrub, grass, savanah
    '111149', # wetlands
    '8dc33b', # croplands
    'cc0013', # urban
    '6ca80d', # crop mosaic
    'd7cdcc', # snow and ice
    'f7e084', # barren
    '6f6f6f'  # tundra
])

map.addLayer(
    image=composite, 
    vis_params={'min':0, 'max':100, 'bands':"30,20,10"}, 
    name='Landsat 7 TOA 2000')
map.addLayer(
    image=cover,
    vis_params={'palette': PALETTE, 'min':0, 'max':17},
    name='MODIS land-cover',
    visible=False)

# Multiple trainings with different numbers of points.
nums = [1000, 10000, 100000]
for num in nums:
    # Extract random points.
    points = ee.FeatureCollection.randomPoints(region, num, num, 1)
    training = cover.addBands(composite).reduceToVectors(
        reducer="mean",
        geometry=points,
        geometryType="centroid",
        scale=30, 
        crs="EPSG:4326"
    )
    
    classifier = training.trainClassifier(
        property_list=bands,
        class_property="label",
        classifier_name="FastNaiveBayes"
    )

    # Apply the classifier to the original composite.
    out = composite.classify(classifier)
    map.addLayer(
        image=out,
        vis_params={'palette': PALETTE, 'min':0, 'max':17},
        name="{0} points".format(num),
        visible=False
    )

# Add the region that constrains the random sampling.
map.addLayer(
        image=ee.Image().paint(ee.Feature(region), 0, 2),
        name="Region"
)


    

In [ ]: