def load_from_pascal_xml_filename(cls, classes, filename, images_dir):
"Load image info from a file in the PASCAL VOC XML format."
def get_data_from_tag(node, tag):
if tag is "bndbox":
x1 = int(node.getElementsByTagName(
tag)[0].childNodes[1].childNodes[0].data)
y1 = int(node.getElementsByTagName(
tag)[0].childNodes[3].childNodes[0].data)
x2 = int(node.getElementsByTagName(
tag)[0].childNodes[5].childNodes[0].data)
y2 = int(node.getElementsByTagName(
tag)[0].childNodes[7].childNodes[0].data)
return (x1, y1, x2, y2)
else:
return node.getElementsByTagName(tag)[0].childNodes[0].data
with open(filename) as f:
data = minidom.parseString(f.read())
# image info
name = get_data_from_tag(data, "filename")
filename = os.path.join(images_dir, name)
size = data.getElementsByTagName("size")[0]
im_width = int(get_data_from_tag(size, "width"))
im_height = int(get_data_from_tag(size, "height"))
width = im_width
height = im_height
img = Image(width, height, classes, name)
# per-object info
objects = []
for obj in data.getElementsByTagName("object"):
clas = str(get_data_from_tag(obj, "name")).lower().strip()
diff = int(get_data_from_tag(obj, "difficult"))
trun = int(get_data_from_tag(obj, "truncated"))
rect = get_data_from_tag(obj, "bndbox")
bbox = BoundingBox(rect, format='corners')
cls_ind = classes.index(clas)
objects.append(np.hstack((bbox.get_arr(), cls_ind, diff, trun)))
if len(objects) > 0:
img.objects_table = Table(np.array(objects), cls.columns)
else:
img.objects_table = Table(None, cls.columns)
return img
def load_from_json(cls, classes, data):
"Return an Image instantiated from a JSON representation."
name = data['name']
width = data['size'][0]
height = data['size'][1]
img = Image(width, height, classes, name)
objects = []
for obj in data['objects']:
bbox = BoundingBox(obj['bbox'])
cls_name = obj['class']
cls_ind = classes.index(cls_name)
diff = obj['diff']
trun = obj['trun']
objects.append(np.hstack((bbox.get_arr(), cls_ind, diff, trun)))
if len(objects) > 0:
img.objects_table = Table(np.array(objects), cls.columns)
else:
img.objects_table = Table(None, cls.columns)
return img
