def test_xml_parse(self):
self.bk.manifest_iter.side_effect = lambda: bk_manifest_iter([(
'media_overlays1', 'file_href1', 'application/smil+xml')])
self.bk.text_iter.side_effect = lambda: bk_text_iter([])
collector = core.XHTMLAttributes()
core.parse_xml(self.bk, collector, self.prefs)
self.assertEqual(collector.class_names, set())
self.assertEqual(collector.literal_class_values, set())
self.assertEqual(collector.id_values, set())
self.assertEqual(
collector.fragment_identifier, {
'ch3_figure1', 'ch3_figure1_title', 'ch3_figure1_caption',
'ch3_figure1_text1', 'ch3_figure1_text2'
})
self.assertEqual(collector.info_class_names, {})
self.assertEqual(collector.info_id_values, {})
def _load_pascal_annotations(_abs_xml_path, _args):
"""
Parse annotations from XML file in Pascal VOC format
:param _abs_xml_path: absolute path of xml file
:param _args: input args
:return: objects, type=dict, for example {'name': [class1, class2, ...], 'area': [area1, area2, ...]}
"""
objects = dict()
objects.setdefault('name', [])
objects.setdefault('area', [])
annotaion = parse_xml(_abs_xml_path)
objs = annotaion['annotation']
if len(objs) == 0:
tqdm.write(" There is no objects in xml : %s" % os.path.basename(_abs_xml_path))
for obj in objs:
objects['name'].append(obj['category'])
objects['area'].append(obj['area'])
return objects
def _load_pascal_annotations(_abs_xml_path, _args):
"""
Parse annotations from XML file in Pascal VOC format
:param _abs_xml_path: absolute path of xml file
:param _args: input args
:return: objects, type=dict, for example {'name': [class1, class2, ...], 'area': [area1, area2, ...]}
"""
objects = dict()
objects.setdefault('name', [])
objects.setdefault('area', [])
#annotaion = parse_xml(_abs_xml_path)
#annotaion=dict()
temp = parse_xml(_abs_xml_path)
if temp is not None:
annotaion = temp
else:
return None
objs = annotaion['annotation']
if len(objs) == 0:
tqdm.write(" There is no objects in xml : %s" %
os.path.basename(_abs_xml_path))
for obj in objs:
#hl->以下四行用来过滤标注不正确的xml文件
# if obj['category'] == 'unknow' or obj['category'] == 'Copy of bus tail' or obj['category'] == 'Copy of car tail' \
# or obj['category'] == 'Copy of car head' or obj['category'] == 'high_60' or obj['category'] == 'high_40':
# print(f"\n{_abs_xml_path} is {obj['category']}")
# return None
objects['name'].append(obj['category'])
objects['area'].append(obj['area'])
return objects
if __name__ == '__main__':
args = parse_args()
src_path = args.org_path
dst_path = args.save_path
xmls = [item for item in os.listdir(src_path) if item.endswith('.xml')]
for idx in tqdm(range(len(xmls)), ncols=100):
xml_name = xmls[idx]
abs_src_xml_path = os.path.join(src_path, xml_name)
abs_dst_xml_path = os.path.join(dst_path, xml_name)
abs_image_path = os.path.join(args.image_path, "%s.png" % os.path.splitext(xml_name)[0])
annotation = parse_xml(abs_src_xml_path)
image_des = annotation['image']
if not os.path.exists(abs_image_path):
tqdm.write("Non existed pic path: %s" % abs_image_path)
continue
cv_image = cv2.imread(abs_image_path)
image_height, image_width= cv_image.shape[:2]
if image_height != image_des['height'] or image_width != image_des['width']:
image_des['height'] = image_height
image_des['width'] = image_width
tqdm.write("Invalid image width/height found in: %s" % xml_name)
def convert(self):
# 遍历所有的XML文件
bnd_id = 1
image_id = 1
for xml_idx in tqdm(range(len(self.m_xml_index)),
ncols=100,
desc="VOC2COCO"):
xml = self.m_xml_index[xml_idx]
abs_xml_path = os.path.join(self.m_xml_path, "%s.xml" % xml)
if not os.path.exists(abs_xml_path):
raise ValueError("Non existed xml path: %s" % abs_xml_path)
annotaion = parse_xml(abs_xml_path, "png")
print(abs_xml_path + '\n')
image_des = annotaion['image']
image_des['id'] = image_id
abs_image_path = os.path.join(self.m_pics_path,
image_des['file_name'])
if not os.path.exists(abs_image_path):
tqdm.write("Non existed pic path: %s" % abs_image_path)
continue
self.m_json_dict['images'].append(image_des)
# TODO: Support segmentation. Currently we do not support segmentation.
objs = annotaion['annotation']
if len(objs) == 0:
tqdm.write(" There is no objects in xml : %s" %
os.path.basename(xml))
for obj in objs:
# 所有标签均以小写格式保存，兼容xml中出现大写字母的情况
category = obj['category'].lower()
if category == 'prohibit':
category = 'limitspeed'
# 排除不需要的标签
if category not in self.m_categories:
continue
'''
if obj['area'] < 40*40:
print('area')
continue
'''
category_id = self.m_categories[category]
obj_annotation = {
'area': obj['area'],
'iscrowd': 0,
'bbox': obj['bbox'],
'category_id': category_id,
'id': bnd_id,
'ignore': 0,
'segmentation': [],
'image_id': image_id
}
self.m_json_dict['annotations'].append(obj_annotation)
bnd_id += 1
image_id += 1
# category
for category, category_id in self.m_categories.items():
#ignore background
if category == 'background':
continue
self.m_json_dict['categories'].append({
'supercategory':
category.split(' ')[0],
'id':
category_id,
'name':
category
})
# save json
with open(self.m_save_path, 'w') as fjson: ##todo
json.dump(self.m_json_dict, fjson)
xml_list = mmcv.list_from_file(args.image_set)
xml_list = list(map(lambda x: x + '.xml', xml_list))
mismatched_nums = 0
# 遍历xml 列表
for i in tqdm(range(len(xml_list)), ncols=100, desc='Merging '):
xml = xml_list[i]
abs_src_xml_path1 = os.path.join(args.anno_path1, xml)
abs_src_xml_path2 = os.path.join(args.anno_path2, xml)
if not (os.path.exists(abs_src_xml_path1)
and os.path.exists(abs_src_xml_path2)):
mismatched_nums += 1
tqdm.write("**** Mismatched xml: %s" % xml)
continue
src_anno1 = parse_xml(abs_src_xml_path1)
src_anno2 = parse_xml(abs_src_xml_path2)
# # check whether image is same
# assert src_anno1['image']['file_name'] == src_anno2['image']['file_name'] \
# and src_anno1['image']['width'] == src_anno2['image']['width'] \
# and src_anno1['image']['height'] == src_anno2['image']['height'], "Mismatched xml %s" % xml
dst_anno = src_anno1.copy()
dst_anno['annotation'] += src_anno2['annotation']
abs_dst_xml_path = os.path.join(args.save_path, xml)
dump_xml(dst_anno, abs_dst_xml_path)
print("There are %d mismatched xmls in total!!!" % mismatched_nums)