def sift_target():
image = Image('../source/section_2_sentinel.tif', 'quickbird')
image.precompute_normalization()
clusters = sift_cluster([image])
window = (slice(100, 125, 1), slice(100, 125, 1))
ubyte = image['gray_ubyte']
features = sift(ubyte[window], clusters)
write_target(features, '../target/sift.nc', 'sift', window)
def test_sift(image):
dataset = Dataset("test/data/target/sift.nc", "r", format="NETCDF4")
target = dataset.variables['sift'][:]
slices = dataset.variables['window'][:]
window = slice(*slices[0:3]), slice(*slices[3:6])
clusters = sift_cluster([image])
win = image['gray_ubyte'][window]
features = sift(win, clusters)
same = target == features
assert same.all()
def create_sift_feature(sat_image: SatelliteImage,
window_sizes,
image_name,
n_clusters=32,
cached=True) -> Sift:
cache_key = "kmeans-sift-{}".format(image_name)
if cached and cached_model_exists(cache_key):
kmeans = load_cached_model(cache_key)
print("Loaded cached kmeans {}".format(cache_key))
else:
print("Computing k-means model")
kmeans = sift_cluster(sat_image, n_clusters=n_clusters)
cache_model(kmeans, cache_key)
feature = Sift(kmeans, windows=(window_sizes))
return feature
for test_image, train_images in cv_train_test_split_images(images):
print("test im:{} train_ims:{}".format(test_image, train_images))
image_name = test_image
mask_full_path = "{base_path}/{image_name}_masked.tif".format(
base_path=base_path, image_name=image_name)
test_image_loaded = load_image(test_image)
cached = True
feature_set = FeatureSet()
for feature_name, feature_scale in feature_names:
feature_scale = tuple((fc, fc) for fc in feature_scale)
if feature_name == "SIFT":
sift_clusters = sift_cluster(map(load_image, train_images))
# for fc in feature_scale:
sift = Sift(sift_clusters, windows=feature_scale)
feature_set.add(sift)
cached = False
if feature_name == "TEXTON":
texton_clusters = texton_cluster(
map(load_image, train_images))
# for fc in feature_scale:
texton = Texton(texton_clusters, windows=feature_scale)
feature_set.add(texton)
cached = False
if feature_name == "PANTEX":
# for fc in feature_scale:
pantex = Pantex(windows=feature_scale)
feature_set.add(pantex)