Skip to main content

Apply Azure on newly uploaded DNV videos (uploaded in January 2023)

Types of defect annotation software

  • There are six main types of annotation style:

    • text in a bounded black box.

      • distance is followed by M.
      • distance is at top left.
      • defect is reported using abbreviation.

      image_with_annotation

      image_without_annotation

    • text in a bounded box.

      • distance is followed by m.
      • distance is at the bottom right.
      • defect is reported using full text with clock position and other value such diameter.

      image_with_annotation_2

      image_without_annotation_2

    • text in white font and not in a box

      • distance is followed by m.
      • distance is at bottom left.
      • defect is reported using full text.
      • the defect is reported to the left of the screen. Have some spaces between the distance

      image_with_annotation_3_1

      image_without_annotation_3

    • text in white font and not in a box

      • distance is followed by m.
      • distance is at bottom left.
      • defect is reported just on top of the distance.

      image_with_annotation_3_2

    • text in black font and not in a box.

      • distance is at the upper left.
      • distance is followed by M.
      • defect is reported using abbreviation.

      image_with_annotation_4

      image_without_annotation_4

    • distance and defect is not shown in the video

      • name of video: 500-xxx.m4v

      image_type_6

Capturing text from images of the new videos using Azure

  • I have tested Azure on 29 images from 001-3648 and 03-APR-004 videos.
  • 001-3648 video has type 3 annotation style (text in white no bounding).
    • Azure could not capture the distance from all 001-3648 images except for one image.
    • Azure was able to capture the defect label correctly for all images.
  • 03-APR-004 video has type 1 annotation style (text in a bounded black box).
    • Azure was able to capture the distance from all the images.
    • Azure was able to capture the defect label correctly for all images.
  • 04-JUN-002 video has type 2 annotation style.
    • distances were captured correctly most of the time. However, they were space between full stop and number and m did not get captured correctly. For example, it captured 44.19m as 44. 19m or 44. 19mo.
    • texts that were represented defects were captured correctly.
  • 05-134-348 video has type 5 annotation style.
    • distance texts were captured correctly. However, we will need to edit regex to distinguish between diameter and distance as both numbers are followed by letter m. Diameter is followed by mm and distance is followed by M.
    • texts that were represented defects were captured correctly.
  • 15XE1915X-397 video has type 4 annotation style.
    • distances were captured correctly.
    • texts that were represented defects were captured correctly.
  • We will need to change the code in the notebook to extract the defect text because the defect is no longer followed by the word obser.