def postprocess(self, inputs: Any, **postprocess_parameters: Dict) -> Any:
model_outputs, images, orig_images = unzip(inputs)
orig_img_shapes = [img.shape for img in orig_images]
output = self.detector.postprocessing(model_outputs, images,
orig_img_shapes,
**postprocess_parameters)
return unzip([output, orig_images])
def postprocess(self, inputs: Any, **postprocess_parameters: Dict) -> Any:
confidences, region_ids, count_lines = self.detector.unzip_predicted(
unzip(inputs))
count_lines = self.detector.custom_count_lines_id_to_all_count_lines(
count_lines)
region_names = self.detector.get_region_labels(region_ids)
return unzip(
[region_ids, region_names, count_lines, confidences, inputs])
def forward(self, inputs: Any, **forward_parameters: Dict) -> Any:
images_bboxs, images = unzip(self.number_plate_localization(inputs, **forward_parameters))
images_points, _ = unzip(self.number_plate_key_points_detection(unzip([images, images_bboxs]),
**forward_parameters))
filled_images = []
for key_ponts, image in zip(images_points, images):
image = image.astype(np.uint8)
for cntr in key_ponts:
cntr = np.array(cntr, dtype=np.int32)
cv2.drawContours(image, [cntr], -1, (0, 0, 0), -1)
filled_images.append(image)
return filled_images
def forward_detection_np(self, images: Any, **forward_parameters: Dict):
orig_img_shapes = [img.shape for img in images]
model_inputs = self.localization_detector.normalize_imgs(
images, **forward_parameters)
localization_detector_output = self.localization_detector.model(
model_inputs)[0]
shape = localization_detector_output.shape
localization_detector_output = localization_detector_output.reshape(
(shape[0], 1, shape[1], shape[2]))
images_target_boxes = self.localization_detector.postprocessing(
localization_detector_output, model_inputs, orig_img_shapes,
**forward_parameters)
images_points, images_mline_boxes = self.key_points_detector.detect(
unzip([images, images_target_boxes]), **forward_parameters)
zones, image_ids = crop_number_plate_zones_from_images(
images, images_points)
zones_model_input = self.option_detector.preprocess(zones)
options_output = self.option_detector.forward(zones_model_input)
options_output = [p.cpu().numpy() for p in options_output]
confidences, region_ids, count_lines = self.option_detector.unzip_predicted(
options_output)
count_lines = self.option_detector.custom_count_lines_id_to_all_count_lines(
count_lines)
region_names = self.option_detector.get_region_labels(region_ids)
return (region_ids, region_names, count_lines, confidences,
options_output, zones, image_ids, images_target_boxes, images,
images_points, images_mline_boxes)
def forward(self, inputs: Any, **forward_parameters: Dict) -> Any:
images_bboxs, images = unzip(self.number_plate_localization(inputs, **forward_parameters))
zones, image_ids = crop_number_plate_rect_zones_from_images(images, images_bboxs)
region_names = [self.text_reader_name for _ in zones]
lines = [self.default_lines_count for _ in zones]
number_plate_text_reading_res = unzip(
self.number_plate_text_reading(unzip([zones,
region_names,
lines]), **forward_parameters))
if len(number_plate_text_reading_res):
texts, _ = number_plate_text_reading_res
else:
texts = []
(texts, zones) = group_by_image_ids(image_ids, (texts, zones))
return unzip([images, images_bboxs,
zones, texts])
def forward_recognition_np(self, region_ids, region_names, count_lines,
confidences, zones, image_ids, images_bboxs,
images, images_points, **forward_parameters):
number_plate_text_reading_res = unzip(
self.number_plate_text_reading(
unzip([zones, region_names, count_lines]),
**forward_parameters))
if len(number_plate_text_reading_res):
texts, _ = number_plate_text_reading_res
else:
texts = []
(region_ids, region_names, count_lines, confidences, texts, zones) = \
group_by_image_ids(image_ids, (region_ids, region_names, count_lines, confidences, texts, zones))
return unzip([
images, images_bboxs, images_points, zones, region_ids,
region_names, count_lines, confidences, texts
])
def forward(self, images: Any, **forward_parameters: Dict) -> Any:
model_inputs = self.detector.normalize_imgs(images,
**forward_parameters)
model_outputs = self.detector.model(model_inputs)[0]
shape = model_outputs.shape
model_outputs = model_outputs.reshape(
(shape[0], 1, shape[1], shape[2]))
return unzip([model_outputs, model_inputs, images])
def detect(data: Dict):
img_path = data['path']
try:
result = number_plate_detection_and_reading([img_path])
(images, images_bboxs, images_points, images_zones, region_ids,
region_names, count_lines, confidences, texts) = unzip(result)
return ujson.dumps(dict(res=texts, img_path=img_path))
except Exception as e:
exc_type, exc_value, exc_tb = sys.exc_info()
traceback.print_exception(exc_type, exc_value, exc_tb)
return ujson.dumps(dict(error=str(e), img_path=img_path))
def forward_detection_np(self, inputs: Any, **forward_parameters: Dict):
images_bboxs, images = unzip(
self.number_plate_localization(inputs, **forward_parameters))
images_points, images_mline_boxes = unzip(
self.number_plate_key_points_detection(
unzip([images, images_bboxs]), **forward_parameters))
zones, image_ids = crop_number_plate_zones_from_images(
images, images_points)
if self.number_plate_classification is None:
region_ids = [-1 for _ in zones]
region_names = [self.default_label for _ in zones]
count_lines = [self.default_lines_count for _ in zones]
confidences = [-1 for _ in zones]
predicted = [-1 for _ in zones]
else:
(region_ids, region_names, count_lines, confidences,
predicted) = unzip(
self.number_plate_classification(zones, **forward_parameters))
return (region_ids, region_names, count_lines, confidences, predicted,
zones, image_ids, images_bboxs, images, images_points,
images_mline_boxes)
def postprocess(self, inputs: Any, **postprocess_parameters: Dict) -> Any:
images_target_boxes, images = unzip(
NumberPlateLocalization.postprocess(self, inputs,
**postprocess_parameters))
filled_number_plate = []
for image_target_boxes, img in zip(images_target_boxes, images):
img = img.astype(np.uint8)
for target_box in image_target_boxes:
cv2.rectangle(img, (int(target_box[0]), int(target_box[1])),
(int(target_box[2]), int(target_box[3])),
(0, 0, 0), -1)
filled_number_plate.append(img)
return filled_number_plate
def read_number_plates(url):
global number_plate_detection_and_reading
with tempfile.NamedTemporaryFile() as fp:
with urlopen(url) as response:
fp.write(response.read())
result = number_plate_detection_and_reading([fp.name])
(images, images_bboxs, images_points, images_zones, region_ids,
region_names, count_lines, confidences, texts) = unzip(result)
return texts[0], region_names[0]
def post(self):
data = tornado.escape.json_decode(self.request.body)
img_path = data['path']
try:
result = number_plate_detection_and_reading([img_path])
(images, images_bboxs, images_points, images_zones, region_ids,
region_names, count_lines, confidences, texts) = unzip(result)
res = ujson.dumps(dict(res=texts, img_path=img_path))
self.write(res)
except Exception as e:
exc_type, exc_value, exc_tb = sys.exc_info()
traceback.print_exception(exc_type, exc_value, exc_tb)
res = ujson.dumps(dict(error=str(e), img_path=img_path))
self.write(res)
def forward(self, inputs: Any, **forward_parameters: Dict) -> Any:
images, labels, lines = unzip(inputs)
model_inputs = self.detector.preprocess(images, labels, lines)
model_outputs = self.detector.forward(model_inputs)
model_outputs = self.detector.postprocess(model_outputs)
return unzip([images, model_outputs, labels])
def postprocess(self, inputs: Any, **postprocess_parameters: Dict) -> Any:
return unzip(
self.detector.postprocess(inputs, **postprocess_parameters))
def forward(self, images: Any, **forward_parameters: Dict) -> Any:
model_inputs = self.detector.normalize_imgs(images, **forward_parameters)
model_outputs = self.detector.forward(model_inputs)
return unzip([model_outputs, model_inputs, images])
import os
from glob import glob
from _paths import nomeroff_net_dir
from nomeroff_net import pipeline
from nomeroff_net.tools import unzip
if __name__ == '__main__':
multiline_number_plate_detection_and_reading = pipeline(
"multiline_number_plate_detection_and_reading", image_loader="opencv")
result = multiline_number_plate_detection_and_reading(
glob(
os.path.join(nomeroff_net_dir,
'./data/examples/multiline_images/*')))
(images, images_bboxs, images_points, images_zones, region_ids,
region_names, count_lines, confidences, texts) = unzip(result)
print(texts)
import os
from _paths import nomeroff_net_dir
from nomeroff_net import pipeline
from nomeroff_net.tools import unzip
if __name__ == '__main__':
number_plate_detection_and_reading = pipeline("number_plate_detection_and_reading", image_loader="opencv")
result = number_plate_detection_and_reading([
os.path.join(nomeroff_net_dir, './data/examples/oneline_images/example1.jpeg'),
])
(images, images_bboxs,
images_points, images_zones, region_ids,
region_names, count_lines,
confidences, texts) = unzip(result)
# (['AC4921CB'], ['RP70012', 'JJF509'])
print(texts)
def postprocess(self, inputs: Any, **postprocess_parameters: Dict) -> Any:
images, model_outputs, labels = unzip(inputs)
outputs = text_postprocessing(model_outputs, labels)
return unzip([outputs, images])
def preprocess(self, inputs: Any, **preprocess_parameters: Dict) -> Any:
images, labels, lines = unzip(inputs)
images = [self.image_loader.load(item) for item in images]
return unzip([images, labels, lines])
def preprocess(self, inputs: Any, **preprocess_parameters: Dict) -> Any:
images = [self.image_loader.load(item) for item in inputs]
batch_input_image, batch_origin_h, batch_origin_w = self.detector.yolov5_wrapper.prepare_batch_input_image(images)
return unzip([images, batch_input_image, batch_origin_h, batch_origin_w])
def forward(self, inputs: Any, **forward_parameters: Dict) -> Any:
model_output = self.detector.forward(inputs)
return unzip([p.cpu().numpy() for p in model_output])
def postprocess(self, inputs: Any, **postprocess_parameters: Dict) -> Any:
images, detected_images_bboxs = unzip(inputs)
detected_images_bboxs = self.detector.postprocessing(detected_images_bboxs, **postprocess_parameters)
return unzip([detected_images_bboxs, images])
def forward(self, inputs: Any, **forward_parameters: Dict) -> Any:
images, batch_input_image, batch_origin_h, batch_origin_w = unzip(inputs)
detected_images_bboxs = self.detector.yolov5_wrapper.infer(batch_input_image, batch_origin_h, batch_origin_w)
return unzip([images, detected_images_bboxs])
