Python生成PASCAL VOC格式的xml标注文件-xml文件

PASCAL VOC数据集的标注文件是xml格式的。对于py-faster-rcnn 或者ssd，通常以下示例的字段是合适的：

1. 
   
2. <annotation>
3. <folder>GTSDB</folder>
4. <filename>000001.webp</filename>
5. <size>
6. <width>500</width>
7. <height>375</height>
8. <depth>3</depth>
9. </size>
10. <object>
11. <name>mouse</name>
12. <difficult>0</difficult>
13. <bndbox>
14. <xmin>99</xmin>
15. <ymin>358</ymin>
16. <xmax>135</xmax>
17. <ymax>375</ymax>
18. </bndbox>
19. </object>
20. </annotation>

怎样从csv或者txt格式的文件，读取bbox信息，生成xml格式的annotation文件呢？直接逐行写文件肯定可以，但是以后改起来并不太方便，\t和空格的替换也不太方便。

1. 
   
2. sudo pip install lxml
5. #from xml.etree.ElementTree import Element, SubElement, tostring
6. from lxml.etree import Element, SubElement, tostring
7. import pprint
8. from xml.dom.minidom import parseString
9. node_root = Element('annotation')
10. node_folder = SubElement(node_root, 'folder')
11. node_folder.text = 'GTSDB'
12. node_filename = SubElement(node_root, 'filename')
13. node_filename.text = '000001.webp'
14. node_size = SubElement(node_root, 'size')
15. node_width = SubElement(node_size, 'width')
16. node_width.text = '500'
17. node_height = SubElement(node_size, 'height')
18. node_height.text = '375'
19. node_depth = SubElement(node_size, 'depth')
20. node_depth.text = '3'
21. node_object = SubElement(node_root, 'object')
22. node_name = SubElement(node_object, 'name')
23. node_name.text = 'mouse'
24. node_difficult = SubElement(node_object, 'difficult')
25. node_difficult.text = '0'
26. node_bndbox = SubElement(node_object, 'bndbox')
27. node_xmin = SubElement(node_bndbox, 'xmin')
28. node_xmin.text = '99'
29. node_ymin = SubElement(node_bndbox, 'ymin')
30. node_ymin.text = '358'
31. node_xmax = SubElement(node_bndbox, 'xmax')
32. node_xmax.text = '135'
33. node_ymax = SubElement(node_bndbox, 'ymax')
34. node_ymax.text = '375'
35. xml = tostring(node_root, pretty_print=True) #格式化显示，该换行的换行
36. dom = parseString(xml)
37. print xml