def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
figure_set = FigureSet()
figure_set.load_list(args.list_path)
with open(args.annotation_path, 'w', newline='') as csvfile:
csv_writer = csv.writer(csvfile, delimiter=',')
for idx, file in enumerate(figure_set.files):
# if '1465-9921-6-55-4' not in file:
# continue
print('Processing Image {:d}: {:s}.'.format(idx, file))
# Create figure object
figure = Figure(file)
# Load image
figure.load_image()
# Load annotation
xml_path = os.path.join(
figure.image_path.replace('.jpg', '_data.xml'))
figure.load_annotation_iphotodraw(xml_path)
# write to CSV file
# The format is:
# image_path,label_x1,label_y1,label_x2,label_y2,label
# if there is no label, the format becomes:
# image_path,,,,,
# TODO : factorize this
has_label = False
for panel in figure.panels:
if panel.label_rect is not None:
label = map_label(panel.label)
if len(label
) == 1: # We care for single letter label for now
row = list()
# add image_path
row.append(figure.image_path)
row.append(str(panel.label_rect[0]))
row.append(str(panel.label_rect[1]))
row.append(str(panel.label_rect[2]))
row.append(str(panel.label_rect[3]))
row.append(label)
csv_writer.writerow(row)
has_label = True
if not has_label:
row = list()
row.append(figure.image_path) # add image_path
row.append('')
row.append('')
row.append('')
row.append('')
row.append('')
csv_writer.writerow(row)
def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
figure_set = FigureSet()
figure_set.load_list(args.list_path)
with open(args.annotation_path, 'w', newline='') as csvfile:
csv_writer = csv.writer(csvfile, delimiter=',')
for idx, file in enumerate(figure_set.files):
# if '1465-9921-6-55-4' not in file:
# continue
print('Processing Image {:d}: {:s}.'.format(idx, file))
figure = Figure(file)
figure.load_image()
xml_path = os.path.join(
figure.image_path.replace('.jpg', '_data.xml'))
figure.load_annotation_iphotodraw(xml_path)
# write to CSV file
# The format is:
# image_path,panel_x1,panel_y1,panel_x2,panel_y2,label_x1,label_y1,label_x2,label_y2,label
# if there is no label, the format becomes:
# image_path,panel_x1,panel_y1,panel_x2,panel_y2,,,,,
for panel in figure.panels:
row = list()
row.append(figure.image_path) # add image_path
row.append(str(panel.panel_rect[0]))
row.append(str(panel.panel_rect[1]))
row.append(str(panel.panel_rect[2]))
row.append(str(panel.panel_rect[3]))
row.append('panel')
if panel.label_rect is None:
row.append('')
row.append('')
row.append('')
row.append('')
row.append('')
else:
label = map_label(panel.label)
row.append(str(panel.label_rect[0]))
row.append(str(panel.label_rect[1]))
row.append(str(panel.label_rect[2]))
row.append(str(panel.label_rect[3]))
row.append(label)
csv_writer.writerow(row)
def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
figure_set = FigureSet()
figure_set.load_list(args.list_path)
with open(args.annotation_path, 'w', newline='') as csvfile:
csv_writer = csv.writer(csvfile, delimiter=',')
for idx, file in enumerate(figure_set.files):
# if '1465-9921-6-55-4' not in file:
# continue
print('Processing Image {:d}: {:s}.'.format(idx, file))
figure = Figure(file)
figure.load_image()
xml_path = os.path.join(figure.image_path.replace('.jpg', '_data.xml'))
figure.load_annotation_iphotodraw(xml_path)
# write to CSV file
# The format is:
# image_path,panel_x1,panel_y1,panel_x2,panel_y2,label_x1,label_y1,label_x2,label_y2,label
# if there is no label, the format becomes:
# image_path,panel_x1,panel_y1,panel_x2,panel_y2,,,,,
for panel in figure.panels:
row = list()
row.append(figure.image_path) # add image_path
row.append(str(panel.panel_rect[0]))
row.append(str(panel.panel_rect[1]))
row.append(str(panel.panel_rect[2]))
row.append(str(panel.panel_rect[3]))
row.append('panel')
if panel.label_rect is None:
row.append('')
row.append('')
row.append('')
row.append('')
row.append('')
else:
label = map_label(panel.label)
row.append(str(panel.label_rect[0]))
row.append(str(panel.label_rect[1]))
row.append(str(panel.label_rect[2]))
row.append(str(panel.label_rect[3]))
row.append(label)
csv_writer.writerow(row)
def to_tf_example(self):
"""
Convert to TfRecord Example
"""
panel_xmins = []
panel_ymins = []
panel_xmaxs = []
panel_ymaxs = []
label_xmins = []
label_ymins = []
label_xmaxs = []
label_ymaxs = []
label_texts = []
label_classes = []
for panel in self.panels:
panel_xmins.append(panel.panel_rect[0] / self.image_width)
panel_ymins.append(panel.panel_rect[1] / self.image_height)
panel_xmaxs.append(panel.panel_rect[2] / self.image_width)
panel_ymaxs.append(panel.panel_rect[3] / self.image_height)
if panel.label_rect is not None:
label_text = panel.label
if len(label_text
) == 1: # We handle 1 char label only for now.
label_text = misc.map_label(label_text)
label_texts.append(label_text.encode('utf8'))
label_classes.append(misc.LABEL_CLASS_MAPPING[label_text])
label_xmins.append(panel.label_rect[0] / self.image_width)
label_ymins.append(panel.label_rect[1] / self.image_height)
label_xmaxs.append(panel.label_rect[2] / self.image_width)
label_ymaxs.append(panel.label_rect[3] / self.image_height)
else:
label_text = misc.map_label(label_text)
label_texts.append(label_text.encode('utf8'))
label_classes.append(
-1) # We handle 1 char label only for now.
label_xmins.append(panel.label_rect[0] / self.image_width)
label_ymins.append(panel.label_rect[1] / self.image_height)
label_xmaxs.append(panel.label_rect[2] / self.image_width)
label_ymaxs.append(panel.label_rect[3] / self.image_height)
else:
label_texts.append('NONE'.encode('utf8'))
label_classes.append(-1)
label_xmins.append(0)
label_xmaxs.append(0)
label_ymaxs.append(0)
label_ymins.append(0)
feature_dict = {
'image/height':
misc.int64_feature(self.image_height),
'image/width':
misc.int64_feature(self.image_width),
'image/filename':
misc.bytes_feature(self.image_path.encode('utf8')),
'image/image':
misc.bytes_feature(tf.compat.as_bytes(self.image.tostring())),
'image/panel/bbox/xmin':
misc.float_list_feature(panel_xmins),
'image/panel/bbox/ymin':
misc.float_list_feature(panel_ymins),
'image/panel/bbox/xmax':
misc.float_list_feature(panel_xmaxs),
'image/panel/bbox/ymax':
misc.float_list_feature(panel_ymaxs),
'image/label/bbox/xmin':
misc.float_list_feature(label_xmins),
'image/label/bbox/ymin':
misc.float_list_feature(label_ymins),
'image/label/bbox/xmax':
misc.float_list_feature(label_xmaxs),
'image/label/bbox/ymax':
misc.float_list_feature(label_ymaxs),
'image/panel/label/text':
misc.bytes_list_feature(label_texts),
'image/panel/label/class':
misc.int64_list_feature(label_classes),
}
example = tf.train.Example(features=tf.train.Features(
feature=feature_dict))
return example
def to_tf_example(self):
"""
Convert to TfRecord Example
"""
panel_xmins = []
panel_ymins = []
panel_xmaxs = []
panel_ymaxs = []
label_xmins = []
label_ymins = []
label_xmaxs = []
label_ymaxs = []
label_texts = []
label_classes = []
for panel in self.panels:
panel_xmins.append(panel.panel_rect[0]/self.image_width)
panel_ymins.append(panel.panel_rect[1]/self.image_height)
panel_xmaxs.append(panel.panel_rect[2]/self.image_width)
panel_ymaxs.append(panel.panel_rect[3]/self.image_height)
if panel.label_rect is not None:
label_text = panel.label
if len(label_text) == 1: # We handle 1 char label only for now.
label_text = misc.map_label(label_text)
label_texts.append(label_text.encode('utf8'))
label_classes.append(misc.LABEL_CLASS_MAPPING[label_text])
label_xmins.append(panel.label_rect[0]/self.image_width)
label_ymins.append(panel.label_rect[1]/self.image_height)
label_xmaxs.append(panel.label_rect[2]/self.image_width)
label_ymaxs.append(panel.label_rect[3]/self.image_height)
else:
label_text = misc.map_label(label_text)
label_texts.append(label_text.encode('utf8'))
label_classes.append(-1) # We handle 1 char label only for now.
label_xmins.append(panel.label_rect[0]/self.image_width)
label_ymins.append(panel.label_rect[1]/self.image_height)
label_xmaxs.append(panel.label_rect[2]/self.image_width)
label_ymaxs.append(panel.label_rect[3]/self.image_height)
else:
label_texts.append('NONE'.encode('utf8'))
label_classes.append(-1)
label_xmins.append(0)
label_xmaxs.append(0)
label_ymaxs.append(0)
label_ymins.append(0)
feature_dict = {
'image/height': misc.int64_feature(self.image_height),
'image/width': misc.int64_feature(self.image_width),
'image/filename': misc.bytes_feature(self.image_path.encode('utf8')),
'image/image': misc.bytes_feature(tf.compat.as_bytes(self.image.tostring())),
'image/panel/bbox/xmin': misc.float_list_feature(panel_xmins),
'image/panel/bbox/ymin': misc.float_list_feature(panel_ymins),
'image/panel/bbox/xmax': misc.float_list_feature(panel_xmaxs),
'image/panel/bbox/ymax': misc.float_list_feature(panel_ymaxs),
'image/label/bbox/xmin': misc.float_list_feature(label_xmins),
'image/label/bbox/ymin': misc.float_list_feature(label_ymins),
'image/label/bbox/xmax': misc.float_list_feature(label_xmaxs),
'image/label/bbox/ymax': misc.float_list_feature(label_ymaxs),
'image/panel/label/text': misc.bytes_list_feature(label_texts),
'image/panel/label/class': misc.int64_list_feature(label_classes),
}
example = tf.train.Example(features=tf.train.Features(feature=feature_dict))
return example