def _read_image_groups(meta_data: MetaDataSet):
image_cluster = {}
cluster_images = defaultdict(list)
for image, cluster in meta_data.load_image_groups():
image_cluster[image] = cluster
cluster_images[cluster].append(image)
K = len(cluster_images)
cluster_index = dict(zip(sorted(cluster_images.keys()), range(K)))
images = []
positions = []
labels = []
centers = np.zeros((K, 2))
centers_count = np.zeros((K, 1))
for image, lat, lon in meta_data.images_with_gps():
images.append(image)
positions.append([lat, lon])
cluster = image_cluster[image]
label = cluster_index[cluster]
labels.append(label)
centers[label, 0] += lat
centers[label, 1] += lon
centers_count[label] += 1
images = np.array(images)
positions = np.array(positions, np.float32)
labels = np.array(labels)
centers /= centers_count
meta_data.save_clusters(images, positions, labels, centers)
def _add_cluster_neighbors(meta_data: MetaDataSet, max_distance):
images, positions, labels, centers = meta_data.load_clusters()
clusters = tools.add_cluster_neighbors(positions, labels, centers,
max_distance)
image_clusters = []
for cluster in clusters:
image_clusters.append(list(np.take(images, np.array(cluster))))
meta_data.save_clusters_with_neighbors(image_clusters)
def _cluster_images(meta_data: MetaDataSet, cluster_size):
images = []
positions = []
for image, lat, lon in meta_data.images_with_gps():
images.append(image)
positions.append([lat, lon])
positions = np.array(positions, np.float32)
images = np.array(images).reshape((len(images), 1))
K = float(images.shape[0]) / cluster_size
K = int(np.ceil(K))
labels, centers = tools.kmeans(positions, K)[1:]
images = images.ravel()
labels = labels.ravel()
meta_data.save_clusters(images, positions, labels, centers)
def _save_clusters_geojson(meta_data: MetaDataSet):
image_coordinates = {}
for image, lat, lon in meta_data.images_with_gps():
image_coordinates[image] = [lon, lat]
features = []
images, positions, labels, centers = meta_data.load_clusters()
for image, label in zip(images, labels):
features.append({
"type": "Feature",
"geometry": {
"type": "Point",
"coordinates": image_coordinates[image]
},
"properties": {
"name": image,
"submodel": int(label)
}, # cluster_idx
})
geojson = {"type": "FeatureCollection", "features": features}
meta_data.save_clusters_geojson(geojson)
def _save_cluster_neighbors_geojson(meta_data: MetaDataSet):
image_coordinates = {}
for image, lat, lon in meta_data.images_with_gps():
image_coordinates[image] = [lon, lat]
features = []
clusters = meta_data.load_clusters_with_neighbors()
for cluster_idx, images in enumerate(clusters):
for image in images:
features.append({
"type": "Feature",
"geometry": {
"type": "Point",
"coordinates": image_coordinates[image],
},
"properties": {
"name": image,
"submodel": cluster_idx
},
})
geojson = {"type": "FeatureCollection", "features": features}
meta_data.save_cluster_with_neighbors_geojson(geojson)
def run_dataset(data: DataSet):
""" Split the dataset into smaller submodels. """
meta_data = MetaDataSet(data.data_path)
meta_data.remove_submodels()
data.invent_reference_lla()
_create_image_list(data, meta_data)
if meta_data.image_groups_exists():
_read_image_groups(meta_data)
else:
_cluster_images(meta_data, data.config["submodel_size"])
_add_cluster_neighbors(meta_data, data.config["submodel_overlap"])
_save_clusters_geojson(meta_data)
_save_cluster_neighbors_geojson(meta_data)
meta_data.create_submodels(meta_data.load_clusters_with_neighbors())