Documentation of the FingerData.foldedness value is incorrect

[Carola_C_]

Hello,

when using the PXCHandData::FingerData.foldedness value, I noticed that the documentation is incorrect. In the manual (page 952) it says

foldedness       The degree of foldedness from 0 (extended) to 100 (folded).

But the value that I get is 0 when a finger is completely folded and 100 when it is completely extended. So either the functions return value should be the other way round or the value should be called "extendedness", then it would be correct.

Also there seems to be an issue with the thumb, most of the time the foldedness value stays 50 or 100 when I fold the thumb and doesn't reach 0.

[Pubudu-Silva_Intel]

Hello Carola,

Hope you are having fun with our beta release. Thanks for brining this to our attention. Let me have a look in to this and get back to you soon.

[Pubudu-Silva_Intel]

Carola,

I finally had sometime to look in to this and you are right! We are reporting the degree of extension although the variable is called 'foldedness'. Great catch!!! This kind of attention to detail will surely help us release a perfect gold version from the current beta.  

About the thumb, I agree it is not very robust. Compared to other four fingers thumb is relatively difficult as it is naturally occluded by the other four fingers. For example, if you turn your hand in a way that the thumb is closer to the camera than others, you will experience more robust behavior.

Again I really appreciate your input and please let us know as you have more feedback.

[Matthew_P_]

To add to this discussion, it would be appreciated if the foldedness value really did range from 0 to 100 depending on how rigid the finger is.  Currently it only returns one of three values: 0, 50, 100 so there is no room to add a threshold in determining how much the finger is extended.

The radius value of the fingerData is also peculiar to me. I'm not sure of its purpose currently as it always returns 0.017 which makes little practical sense. I felt it should instead return the approximate pixel size of the detected finger tip based on distance from camera.

[Dagan_E_Intel]

Hi Carola and Matthew,

Thanks a lot for your valuable comments.

The calculation of the foldedness values was modified and in the next release it will be much more accurate, and cover the full range from 0 to 100. Also the thumb was improved, although, as stated before, it is a more difficult case.

Look at your thumb when folding it, or folding the other fingers and you will see the problems :)

Regarding the finger radius.

This value depends on a successful calibration of the hand.

Once the hand is calibrated (you can get an alert for that) the radius value will change accordingly.

The calibration can be done faster if you extend your hand in front of the camera (for example)

hth,

Dagan

[Scott_W_1]

Has the recalculation been released? I have an app almost ready for distribution. It records a single-frame hand pose so people can eventually add new gestures. I could work around the inverted foldedness values, but would like to use the updated SDK if it has the new values already set.

I'm using the hand_viewer sample that came with the SDK sample browser.