def load_convnet(self):
self.convnet = Detector(labels_file=self.configuration.labels,
graph_file=self.configuration.network)
if not self.configuration.word in self.convnet.labels:
error_message = self.configuration.word + " is not in label file."
logger.error(error_message)
raise Exception(error_message)
def main_function():
parser = argparse.ArgumentParser(
description='Test a trained neuronal network.', )
parser.add_argument('-i',
'--image',
action='store',
dest='image',
required=True,
help='The path to the input image file.')
parser.add_argument('-g',
'--graph',
action='store',
dest='graph',
required=True,
help='The path to the graph file.')
parser.add_argument('-l',
'--label',
action='store',
dest='label',
required=True,
help='The path to the label file.')
args = parser.parse_args()
for key, value in vars(args).items():
check_file(value)
detector = Detector(graph_file=args.graph, labels_file=args.label)
image = Image.open(args.image)
answer = detector.detect([image])
print(answer)
class BoxWalker:
def __init__(self, bbox, configuration=None):
self.configuration = Configuration(
) if configuration is None else configuration
self.bbox = bbox
self.tile = None
self.streets = []
self.convnet = None
self.square_image_length = 50
self.max_distance = self._calculate_max_distance(
self.configuration.zoom_level, self.square_image_length)
self.image_api = OtherApi(
self.configuration.zoom_level
) if self.configuration.orthophoto is 'other' else self._get_image_api(
self.configuration.orthophoto)
@staticmethod
def _get_image_api(image_api):
source = 'src.data.orthofoto.' + image_api + '.' + image_api + '_api'
module = import_module(source)
class_ = getattr(module, image_api.title() + 'Api')
return class_()
def load_convnet(self):
self.convnet = Detector(labels_file=self.configuration.labels,
graph_file=self.configuration.network)
if not self.configuration.word in self.convnet.labels:
error_message = self.configuration.word + " is not in label file."
logger.error(error_message)
raise Exception(error_message)
def load_tiles(self):
self._printer("Start image loading.")
loader = TileLoader(bbox=self.bbox, image_api=self.image_api)
loader.load_tile()
self.tile = loader.tile
self._printer("Stop image loading.")
def load_streets(self):
self._printer("Start street loading.")
street_loader = StreetLoader()
self.streets = street_loader.load_data(self.bbox)
self._printer(str(len(self.streets)) + " streets to walk.")
self._printer("Stop street loading.")
def walk(self):
ready_for_walk = (not self.tile is None) and (not self.convnet is None)
if not ready_for_walk:
error_message = "Not ready for walk. Load tiles and convnet first"
logger.error(error_message)
raise Exception(error_message)
self._printer("Start detection.")
if self.configuration.follow_streets:
tiles = self._get_tiles_of_box_with_streets(
self.streets, self.tile)
else:
tiles = self._get_tiles_of_box(self.tile)
self._printer(str(len(tiles)) + " images to analyze.")
images = [tile.image for tile in tiles]
predictions = self.convnet.detect(images)
detected_nodes = []
for i, _ in enumerate(tiles):
prediction = predictions[i]
if self.hit(prediction):
detected_nodes.append(tiles[i].get_centre_node())
self._printer("Stop detection.")
merged_nodes = self._merge_near_nodes(detected_nodes)
if self.configuration.compare:
return self._compare_with_osm(merged_nodes)
return merged_nodes
def _get_tiles_of_box(self, tile):
tile_walker = TileWalker(tile=tile,
square_image_length=self.square_image_length,
zoom_level=self.configuration.zoom_level)
tiles = tile_walker.get_tiles()
return tiles
def _get_tiles_of_box_with_streets(self, streets, tile):
street_walker = StreetWalker(
tile=tile,
square_image_length=self.square_image_length,
zoom_level=self.configuration.zoom_level)
tiles = []
for street in streets:
street_tiles = street_walker.get_tiles(street)
tiles += street_tiles
return tiles
def _merge_near_nodes(self, node_list):
merger = NodeMerger(node_list, self.max_distance)
return merger.reduce()
def _compare_with_osm(self, detected_nodes):
comparator = OsmComparator(max_distance=self.max_distance)
return comparator.compare(detected_nodes=detected_nodes,
tag=self.configuration.tag,
bbox=self.bbox)
def hit(self, prediction):
return prediction[self.configuration.word] > self.configuration.barrier
@staticmethod
def _printer(message):
print(str(datetime.datetime.now()) + ": " + message)
@staticmethod
def _calculate_max_distance(zoom_level, square_image_length):
global_mercator = GlobalMercator()
resolution = global_mercator.Resolution(zoom_level)
return resolution * (square_image_length / 2)
def load_convnet(self):
self.convnet = Detector(labels_file=self.configuration.labels, graph_file=self.configuration.network)
if not self.configuration.word in self.convnet.labels:
error_message = self.configuration.word + " is not in label file."
logger.error(error_message)
raise Exception(error_message)
class BoxWalker:
def __init__(self, bbox, configuration=None):
self.configuration = Configuration() if configuration is None else configuration
self.bbox = bbox
self.tile = None
self.streets = []
self.convnet = None
self.square_image_length = 50
self.max_distance = self._calculate_max_distance(self.configuration.zoom_level, self.square_image_length)
self.image_api = OtherApi(
self.configuration.zoom_level) if self.configuration.orthophoto is 'other' else self._get_image_api(
self.configuration.orthophoto)
@staticmethod
def _get_image_api(image_api):
source = 'src.data.orthofoto.' + image_api + '.' + image_api + '_api'
module = import_module(source)
class_ = getattr(module, image_api.title() + 'Api')
return class_()
def load_convnet(self):
self.convnet = Detector(labels_file=self.configuration.labels, graph_file=self.configuration.network)
if not self.configuration.word in self.convnet.labels:
error_message = self.configuration.word + " is not in label file."
logger.error(error_message)
raise Exception(error_message)
def load_tiles(self):
self._printer("Start image loading.")
loader = TileLoader(bbox=self.bbox, image_api=self.image_api)
loader.load_tile()
self.tile = loader.tile
self._printer("Stop image loading.")
def load_streets(self):
self._printer("Start street loading.")
street_loader = StreetLoader()
self.streets = street_loader.load_data(self.bbox)
self._printer(str(len(self.streets)) + " streets to walk.")
self._printer("Stop street loading.")
def walk(self):
ready_for_walk = (not self.tile is None) and (not self.convnet is None)
if not ready_for_walk:
error_message = "Not ready for walk. Load tiles and convnet first"
logger.error(error_message)
raise Exception(error_message)
self._printer("Start detection.")
if self.configuration.follow_streets:
tiles = self._get_tiles_of_box_with_streets(self.streets, self.tile)
else:
tiles = self._get_tiles_of_box(self.tile)
self._printer(str(len(tiles)) + " images to analyze.")
images = [tile.image for tile in tiles]
predictions = self.convnet.detect(images)
detected_nodes = []
for i, _ in enumerate(tiles):
prediction = predictions[i]
if self.hit(prediction):
detected_nodes.append(tiles[i].get_centre_node())
self._printer("Stop detection.")
merged_nodes = self._merge_near_nodes(detected_nodes)
if self.configuration.compare:
return self._compare_with_osm(merged_nodes)
return merged_nodes
def _get_tiles_of_box(self, tile):
tile_walker = TileWalker(tile=tile, square_image_length=self.square_image_length,
zoom_level=self.configuration.zoom_level)
tiles = tile_walker.get_tiles()
return tiles
def _get_tiles_of_box_with_streets(self, streets, tile):
street_walker = StreetWalker(tile=tile, square_image_length=self.square_image_length,
zoom_level=self.configuration.zoom_level)
tiles = []
for street in streets:
street_tiles = street_walker.get_tiles(street)
tiles += street_tiles
return tiles
def _merge_near_nodes(self, node_list):
merger = NodeMerger(node_list, self.max_distance)
return merger.reduce()
def _compare_with_osm(self, detected_nodes):
comparator = OsmComparator(max_distance=self.max_distance)
return comparator.compare(detected_nodes=detected_nodes, tag=self.configuration.tag, bbox=self.bbox)
def hit(self, prediction):
return prediction[self.configuration.word] > self.configuration.barrier
@staticmethod
def _printer(message):
print(str(datetime.datetime.now()) + ": " + message)
@staticmethod
def _calculate_max_distance(zoom_level, square_image_length):
global_mercator = GlobalMercator()
resolution = global_mercator.Resolution(zoom_level)
return resolution * (square_image_length / 2)
def detector():
return Detector()
def load_convnet(self):
self.convnet = Detector()
if not self.search.word in self.convnet.labels:
error_message = self.search.word + " is not in label file."
self.logger.error(error_message)
raise Exception(error_message)
