Format detection fails to disambiguate labelme and voc on tfds voc2012

[zhiltsov-max]

When I try to detect format of a tfds/voc2012, Datumaro detects both LabelMe and VOC. Probably, the problem is in XMLs with only image info inside, like this:

<annotation>
  <folder>2008</folder>
  <filename>2008_000001.jpg</filename>
  <source>
    <database>Unknown</database>
    <annotation>Unknown</annotation>
    <image>Unknown</image>
  </source>
  <segmented>0</segmented>
</annotation>