You are on page 1of 8

Questions about Bio ICAO SDK

The tests performed in this material were performed based on the photos
have already available on the standard ICAO

DOC 9303 Part 1 Vol 1 Item 4. Compliance with international standard
sub-item 4.1 the photographic shall comply with the appropriate
definitions set out in ISO/IEC 19794-5 pages 95 to 97 Portrait Quality,
Style and Lighting Glasses and Head Covers, Expressions and Frame.

The items that we do not understand or that showed divergence will be
marked with squares or red lines.

First Case
We dont understand the items marked in red here because the values presented are low but
the photo shows: no blur, and in the analysis of glasses the values do not match the picture, as
can be seen in the photo below.
Obs.: According icao standard this photo is ok.

Answer: Glasses analisys has been improved. Please note that image quality is not the best:
there are some image jittering (usually due to jpeg compression) that could have effected the
"Blurred" analisys.


Second Case
The SDK shows that the picture has glasses on the eyes, but it does not match the image.

Answer: Glasses are so close to eyes and the frame is so heavy that they are considered over
eyes.
Third Case
In this case the image's mouth is fully open and the score of the SDK was given as 100, and also
that the eyes are fully open and the score was 68.

Answer: A score of 68 for "Eyes closed" means eyes are ok. We can work on mouth score the
next week.

Fourth Case
In this case the mouth was incorrectly detected as image below, and scores are also
divergent.

Answer: Please note that this photo is very bluish. When the color or light is not ok, you can
have the detailed information of what is wrong looking at the structure BioFace returned by
BioICAO_GetFace(). In this case colorOffsetBlue is very high.
The right skin colour is as in the following image, and more close to the image in the third case.








Fifth Case
In this case the SDK showed divergent scores of mouth and
rotation.

Answer: The nose wasn't located, so on this image the rotation wasn't calculated well.

Sixth Case
Glasses score divergent.

Answer: Usually those eyes are considered too closed.
The image is strongly washed out, so glasses analisys are less reliable. A new SDK version is
currently under test that improves glasses detection.
Seventh Case
In this case analysis of glasses was
made.

Answer: A new SDK version is currently under test that improves glasses detection.

Eighth Case
There are many cases where the mouth is incorrectly marked, and the analysis is not
ok.

Answer: the next week we can work on improving mouth detection.

Ninth Case

Answer: eyse are ok, and the score is >60. Please note that our SDK doesn't check the
presence of two faces.

Tenth Case

Answer: the algorithm can be wrong sometimes, expecially when there are other ICAO
compliancy problems like blurred image. Also here the image is bluish.


Eleventh Case

Answer: the next week we can work on improving mouth detection.
Twelfth Case
Obs.: According icao standard this photo is ok.

Answer: this picture is ok for the veil/hat/cover issue, but it is very
bluish.
This on the right is a more correct color. Usually this problem is due to
a wrong set of the white balance of the camera, and the operator
should be warned to check it.


Thirteenth
case

Answer: please note our SDK doesn't check for hats.

You might also like