def setUpClass(cls):
cls.test_cases = [
cls.InterpolationTestCase('linear', LinearInterpolation('NDVI', result_interval=(0.0, 1.0),
mask_feature=(FeatureType.MASK, 'IS_VALID'),
unknown_value=10),
result_len=68, img_min=0.0, img_max=10.0, img_mean=0.720405,
img_median=0.59765935),
cls.InterpolationTestCase('linear_change_timescale',
LinearInterpolation('NDVI', result_interval=(0.0, 1.0),
mask_feature=(FeatureType.MASK, 'IS_VALID'),
unknown_value=10, scale_time=1),
result_len=68, img_min=0.0, img_max=10.0, img_mean=0.720405,
img_median=0.597656965),
cls.InterpolationTestCase('cubic', CubicInterpolation('NDVI', result_interval=(0.0, 1.0),
mask_feature=(FeatureType.MASK, 'IS_VALID'),
resample_range=('2015-01-01', '2018-01-01', 16),
unknown_value=5, bounds_error=False),
result_len=69, img_min=0.0, img_max=5.0, img_mean=1.3592644,
img_median=0.6174331),
cls.InterpolationTestCase('spline', SplineInterpolation('NDVI', result_interval=(-0.3, 1.0),
mask_feature=(FeatureType.MASK, 'IS_VALID'),
resample_range=('2016-01-01', '2018-01-01', 5),
spline_degree=3, smoothing_factor=0,
unknown_value=0),
result_len=147, img_min=-0.3, img_max=1.0, img_mean=0.492752,
img_median=0.53776133),
cls.InterpolationTestCase('bspline', BSplineInterpolation('NDVI', unknown_value=-3,
mask_feature=(FeatureType.MASK, 'IS_VALID'),
resample_range=('2017-01-01', '2017-02-01', 50),
spline_degree=5),
result_len=1, img_min=-0.032482587, img_max=0.701796, img_mean=0.42080238,
img_median=0.42889267),
cls.InterpolationTestCase('akima', AkimaInterpolation('NDVI', unknown_value=0,
mask_feature=(FeatureType.MASK, 'IS_VALID')),
result_len=68, img_min=-0.13793105, img_max=0.860242, img_mean=0.53159297,
img_median=0.59087014),
cls.InterpolationTestCase('kriging interpolation',
KrigingInterpolation('NDVI', result_interval=(-10, 10),
resample_range=('2016-01-01', '2018-01-01', 5)),
result_len=147, img_min=-0.252500534, img_max=0.659086704, img_mean=0.3825493,
img_median=0.39931053),
cls.InterpolationTestCase('nearest resample', NearestResampling('NDVI', result_interval=(0.0, 1.0),
resample_range=('2016-01-01',
'2018-01-01', 5)),
result_len=147, img_min=-0.2, img_max=0.860242, img_mean=0.35143828,
img_median=0.37481314, nan_replace=-0.2),
cls.InterpolationTestCase('linear resample', LinearResampling('NDVI', result_interval=(0.0, 1.0),
resample_range=('2016-01-01',
'2018-01-01', 5)),
result_len=147, img_min=-0.2, img_max=0.8480114, img_mean=0.350186,
img_median=0.3393997, nan_replace=-0.2),
cls.InterpolationTestCase('cubic resample', CubicResampling('NDVI', result_interval=(-0.2, 1.0),
resample_range=('2015-01-01', '2018-01-01', 16),
unknown_value=5),
result_len=69, img_min=-0.2, img_max=5.0, img_mean=1.234881997,
img_median=0.465670556, nan_replace=-0.2),
cls.InterpolationTestCase('linear custom list', LinearInterpolation(
'NDVI', result_interval=(-0.2, 1.0),
resample_range=('2015-09-01', '2016-01-01', '2016-07-01', '2017-01-01', '2017-07-01'),
unknown_value=-2),
result_len=5, img_min=-0.032482587, img_max=0.8427637, img_mean=0.5108417,
img_median=0.5042224),
cls.InterpolationTestCase('linear with bands and multiple masks',
LinearInterpolation('BANDS-S2-L1C', result_interval=(0.0, 1.0), unknown_value=10,
mask_feature=[(FeatureType.MASK, 'IS_VALID'),
(FeatureType.MASK_TIMELESS, 'RANDOM_UINT8'),
(FeatureType.LABEL, 'RANDOM_DIGIT')]),
result_len=68, img_min=0.000200, img_max=10.0, img_mean=0.34815648,
img_median=0.1003600)
]
cls.copy_feature_cases = [
cls.InterpolationTestCase('cubic_copy_success',
CubicInterpolation('NDVI', result_interval=(0.0, 1.0),
mask_feature=(
FeatureType.MASK, 'IS_VALID'),
resample_range=(
'2015-01-01', '2018-01-01', 16),
unknown_value=5, bounds_error=False,
copy_features=[
(FeatureType.MASK, 'IS_VALID'),
(FeatureType.DATA, 'NDVI', 'NDVI_OLD'),
(FeatureType.MASK_TIMELESS, 'LULC'),
]),
result_len=69, img_min=0.0, img_max=5.0, img_mean=1.3592644,
img_median=0.6174331),
cls.InterpolationTestCase('cubic_copy_fail',
CubicInterpolation('NDVI', result_interval=(0.0, 1.0),
mask_feature=(FeatureType.MASK, 'IS_VALID'),
resample_range=(
'2015-01-01', '2018-01-01', 16),
unknown_value=5, bounds_error=False,
copy_features=[
(FeatureType.MASK, 'IS_VALID'),
(FeatureType.DATA, 'NDVI'),
(FeatureType.MASK_TIMELESS, 'LULC'),
]),
result_len=69, img_min=0.0, img_max=5.0, img_mean=1.3592644,
img_median=0.6174331),
]
for test_case in cls.test_cases:
test_case.execute()
for test_case in cls.copy_feature_cases:
try:
test_case.execute()
except ValueError:
pass
# TASK FOR CONCATENATION
concatenate = ConcatenateData('FEATURES', ['BANDS', 'NDVI', 'NDWI', 'NORM'])
# TASK FOR FILTERING OUT TOO CLOUDY SCENES
# keep frames with > 80 % valid coverage
valid_data_predicate = ValidDataFractionPredicate(0.8)
filter_task = SimpleFilterTask((FeatureType.MASK, 'IS_VALID'),
valid_data_predicate)
# TASK FOR LINEAR INTERPOLATION
# linear interpolation of full time-series and date resampling
resampled_range = ('2017-01-01', '2017-12-31', 16)
linear_interp = LinearInterpolation(
'FEATURES', # name of field to interpolate
mask_feature=(FeatureType.MASK,
'IS_VALID'), # mask to be used in interpolation
copy_features=[(FeatureType.MASK_TIMELESS, 'LULC')], # features to keep
resample_range=resampled_range, # set the resampling range
bounds_error=False # extrapolate with NaN's
)
# TASK FOR EROSION
# erode each class of the reference map
erosion = ErosionTask(mask_feature=(FeatureType.MASK_TIMELESS, 'LULC',
'LULC_ERODED'),
disk_radius=1)
# TASK FOR SPATIAL SAMPLING
# Uniformly sample about pixels from patches
n_samples = int(1e5) # no. of pixels to sample
ref_labels = [0, 1, 2, 3, 4,
5] # reference labels to take into account when sampling
# ndvi
ndvi_task = NormalizedDifferenceIndexTask((FeatureType.DATA, 'S2-BANDS-L2A'),
(FeatureType.DATA, 'NDVI'),
[7,3])
# interpolation
lin_interp = LinearInterpolation((FeatureType.DATA, 'S2-BANDS-L2A'),
mask_feature=(FeatureType.MASK, 'VALID_DATA'),
copy_features=[(FeatureType.DATA, 'NDVI'),
(FeatureType.DATA, 'FCI'),
(FeatureType.DATA, 'ARVI'),
(FeatureType.DATA, 'BSI'),
(FeatureType.DATA, 'IAD'),
(FeatureType.DATA, 'CTVI'),
(FeatureType.DATA, 'GEMI'),
(FeatureType.DATA, 'WDRVI'),
(FeatureType.DATA, 'EVI'),
(FeatureType.DATA, 'MTVI'),
(FeatureType.DATA, 'CARI2'),
(FeatureType.DATA, 'ATSAVI'),
(FeatureType.DATA, 'GNDVI'),
(FeatureType.DATA, 'MSR'),
(FeatureType.DATA, 'NVI'),
(FeatureType.DATA, 'RDVI')],
resample_range=dates)
# add false color infrared feature
false_color_infrared_task = ZipFeatureTask({FeatureType.DATA: ['B08', 'B04', 'B03']},
(FeatureType.DATA, 'FCI'),
false_color_infrared)
# layer='BANDS-S2-L1C',
# feature=(FeatureType.DATA, 'BANDS'), # save under name 'BANDS'
# resx='10m', # resolution x
# resy='10m', # resolution y
# maxcc=0.8, # maximum allowed cloud cover of original ESA tiles
# )
cloud_classifier = get_s2_pixel_cloud_detector(average_over=2, dilation_size=1, all_bands=False)
add_clm = AddCloudMaskTask(cloud_classifier, 'BANDS-S2CLOUDLESS', cm_size_y='80m', cm_size_x='80m',
cmask_feature='CLM', # cloud mask name
cprobs_feature='CLP' # cloud prob. map name
)
linear_interp = LinearInterpolation(
(FeatureType.DATA, 'BANDS'), # name of field to interpolate
mask_feature=(FeatureType.MASK, 'IS_VALID'), # mask to be used in interpolation
bounds_error=False # extrapolate with NaN's
)
add_sh_valmask = AddValidDataMaskTask(SentinelHubValidData(),
'IS_VALID' # name of output mask
)
execution_args = []
for id in broken_patches:
execution_args.append({
load: {'eopatch_folder': 'eopatch_{}'.format(id)},
save: {'eopatch_folder': 'eopatch_{}'.format(id)}
})
structuring_2d = [[0, 1, 0],
count_fields,
train=False)
# define a valid data mask used in interpolation
valid_data = MapFeatureTask((FeatureType.MASK, 'CLOUD'),
(FeatureType.MASK, 'VALID_DATA'),
np.logical_not)
# ndvi
ndvi_task = NormalizedDifferenceIndexTask((FeatureType.DATA, 'S2-BANDS-L2A'),
(FeatureType.DATA, 'NDVI'),
[7,3])
# interpolation
lin_interp = LinearInterpolation((FeatureType.DATA, 'S2-BANDS-L2A'),
mask_feature=(FeatureType.MASK, 'VALID_DATA'),
copy_features=[(FeatureType.DATA, 'NDVI')], # comment this out if NDVI is not used
resample_range=dates)
# sample only fields
sample = SampleValid((FeatureType.MASK_TIMELESS, 'FIELD_ID'), fraction=1.0, no_data_value=0)
# if invalid data are at beginning or end of time-series, pad with nearest value
fillout = ValueFilloutTask((FeatureType.DATA, 'S2-BANDS-L2A'))
workflow = ComposeTask([cloud_mask,
#cloud_erosion,
cloud_dilation,
add_cc,
#filter_task,
count_train_fields,
count_test_fields])
def setUpClass(cls):
cls.test_cases = [
cls.InterpolationTestCase('linear',
LinearInterpolation(
'ndvi',
result_interval=(-0.3, -0.1),
unknown_value=10),
result_len=180,
img_min=-0.3000,
img_max=10.0,
img_mean=-0.147169,
img_median=-0.21132),
cls.InterpolationTestCase(
'cubic',
CubicInterpolation('ndvi',
result_interval=(-0.3, -0.1),
resample_range=('2015-01-01', '2018-01-01',
16),
unknown_value=5),
result_len=69,
img_min=-0.3000,
img_max=5.0,
img_mean=1.30387,
img_median=-0.1007357),
cls.InterpolationTestCase(
'spline',
SplineInterpolation('ndvi',
result_interval=(-10, 10),
resample_range=('2016-01-01', '2018-01-01',
5),
spline_degree=3,
smoothing_factor=0),
result_len=147,
img_min=-0.68187547,
img_max=0.26355705,
img_mean=-0.1886001,
img_median=-0.18121514),
cls.InterpolationTestCase('bspline',
BSplineInterpolation(
'ndvi',
unknown_value=-3,
resample_range=('2017-01-01',
'2017-02-01', 50),
spline_degree=5),
result_len=1,
img_min=-0.20366311,
img_max=0.035470814,
img_mean=-0.04307341,
img_median=-0.040708482),
cls.InterpolationTestCase('akima',
AkimaInterpolation('ndvi',
unknown_value=0),
result_len=180,
img_min=-0.4821199,
img_max=0.2299331,
img_mean=-0.20141865,
img_median=-0.213559),
cls.InterpolationTestCase(
'nearest resample',
NearestResampling('ndvi',
result_interval=(-0.3, -0.1),
resample_range=('2016-01-01', '2018-01-01',
5)),
result_len=147,
img_min=-0.3000,
img_max=-0.1000,
img_mean=-0.19651729,
img_median=-0.20000,
nan_replace=-0.2),
cls.InterpolationTestCase(
'linear resample',
LinearResampling('ndvi',
result_interval=(-0.3, -0.1),
resample_range=('2016-01-01', '2018-01-01',
5)),
result_len=147,
img_min=-0.3000,
img_max=-0.1000,
img_mean=-0.1946315,
img_median=-0.20000,
nan_replace=-0.2),
cls.InterpolationTestCase(
'cubic resample',
CubicResampling('ndvi',
result_interval=(-0.3, -0.1),
resample_range=('2015-01-01', '2018-01-01',
16),
unknown_value=5),
result_len=69,
img_min=-0.3000,
img_max=5.0,
img_mean=1.22359,
img_median=-0.181167,
nan_replace=-0.2)
]
for test_case in cls.test_cases:
test_case.execute()
def setUpClass(cls):
cls.test_cases = [
cls.InterpolationTestCase('linear',
LinearInterpolation(
'ndvi',
result_interval=(-0.3, -0.1),
unknown_value=10),
result_len=180,
img_min=-0.3000,
img_max=10.0,
img_mean=-0.147169,
img_median=-0.21132),
cls.InterpolationTestCase(
'cubic',
CubicInterpolation('ndvi',
result_interval=(-0.3, -0.1),
resample_range=('2015-01-01', '2018-01-01',
16),
unknown_value=5),
result_len=69,
img_min=-0.3000,
img_max=5.0,
img_mean=1.30387,
img_median=-0.1007357),
cls.InterpolationTestCase(
'spline',
SplineInterpolation('ndvi',
result_interval=(-10, 10),
resample_range=('2016-01-01', '2018-01-01',
5),
spline_degree=3,
smoothing_factor=0.5),
result_len=147,
img_min=-0.431703,
img_max=0.116959,
img_mean=-0.1948325,
img_median=-0.1922528),
cls.InterpolationTestCase('bspline',
BSplineInterpolation(
'ndvi',
unknown_value=-3,
resample_range=('2017-01-01',
'2017-02-01', 50),
spline_degree=5),
result_len=1,
img_min=-0.161574,
img_max=-0.0116416,
img_mean=-0.085181,
img_median=-0.0856696),
cls.InterpolationTestCase('akima',
AkimaInterpolation('ndvi',
unknown_value=0),
result_len=180,
img_min=-0.4821199,
img_max=0.2299331,
img_mean=-0.20141865,
img_median=-0.213559),
cls.InterpolationTestCase(
'linear resample',
LinearResampling('ndvi',
result_interval=(-0.3, -0.1),
resample_range=('2016-01-01', '2018-01-01',
5)),
result_len=147,
img_min=-0.3000,
img_max=-0.1000,
img_mean=-0.1946315,
img_median=-0.20000,
nan_replace=-0.2),
cls.InterpolationTestCase(
'cubic resample',
CubicResampling('ndvi',
result_interval=(-0.3, -0.1),
resample_range=('2015-01-01', '2018-01-01',
16),
unknown_value=5),
result_len=69,
img_min=-0.3000,
img_max=5.0,
img_mean=1.22359,
img_median=-0.181167,
nan_replace=-0.2)
]
for test_case in cls.test_cases:
test_case.execute()