My pareidolia knows no bounds.

More
10 years 10 months ago #21532 by Larry Burford
Hovercraft (at least the real ones we can build now) can't get that high off the ground.

You'd need tall crane or a helicopter.

Please Log in or Create an account to join the conversation.

More
10 years 10 months ago #24347 by rderosa
Replied by rderosa on topic Reply from Richard DeRosa
Woo Hoo! Three hits! What that proves is that a man-made (Martian-made)sculpted face <b>could be</b> detected by this camera's face recognition routine.


File Attachment:


I suspect there's a "depth" component to the algorithm, besides the necessary features. Possibly some kind of gradient measurement.

rd

Please Log in or Create an account to join the conversation.

More
10 years 10 months ago #21894 by Larry Burford
And no "sunglasses"

Please Log in or Create an account to join the conversation.

More
10 years 10 months ago #22002 by Larry Burford
Except for Teddy?

Please Log in or Create an account to join the conversation.

More
10 years 10 months ago #21533 by rderosa
Replied by rderosa on topic Reply from Richard DeRosa
<blockquote id="quote"><font size="2" face="Verdana, Arial, Helvetica" id="quote">quote:<hr height="1" noshade id="quote"><i>Originally posted by Larry Burford</i>
<br />And no "sunglasses"
<hr height="1" noshade id="quote"></blockquote id="quote"></font id="quote">
True. Hmmm, maybe they really put a lot of emphasis on the eyes? Maybe I'll play around with some rudimentary drawings. See if I can draw a face that it captures.

rd

Please Log in or Create an account to join the conversation.

More
10 years 10 months ago #21534 by rderosa
Replied by rderosa on topic Reply from Richard DeRosa
For anyone who is interested, this article pretty much explains face detection algorithms in cameras today. Incidentally, mine was one of the first cameras to have it (mid-2006), so it's not surprising that newer ones can detect profile (or partial profiles), whereas mine can't.

www.photoreview.com.au/tips/Insider-How-...amera-Face-Detection

Some highlights:

<ul><li>As its name implies, face detection is all about finding faces in a scene - and that involves shape recognition. </li><li> Identifying regions in the subject that contain skin color, using a skin filter which detects colour and texture. </li><li>Scanning these areas for patterns that could represent eyes and eyebrows, nostrils, and mouth. </li><li>In the early days of the technology, the algorithms used for the analysis could only detect faces looking directly at the camera. (like mine) </li><li>The latest algorithms can also detect faces in semi-profile. This requires the use of a larger number of detection points and much more processing capability </li></ul>

System Failures:

<ul><li>Inability to identify pets (and other animals). Face detection systems have been programmed to recognise human faces only. </li><li>They will not focus on cat, dog, horse or other animal faces. </li><li>Unwanted detection of faces in works of art. Most face detection systems will automatically focus on faces in paintings and photographs as well as statues. </li><li>Distance-based failures. Face detection systems will not identify faces that are too close to the camera - or too far away </li><li>Because they rely on pattern recognition, they can only identify an entire face so, if the camera 'sees' only one eye and part of the subject's nose, detection could fail. </li><li>Sunglasses, swimming goggles and other things that cover the eyes can cause face detection systems to fail. </li><li>False detections. These are becoming less common as algorithms are refined. However, there are situations where the face detection system will incorrectly identify patterns on walls, fabrics or similarly-complex subjects as faces. (computer pareidolia? - modern) </li></ul>


Conclusions:

<ul><li>I need a new camera for Christmas. One that is state of the art.</li><li>The failures may be as interesting as the successes, since that's what we're talking about here. </li><li>Since skin color is a major feature, it may totally rule out most known pareidolia (modern or ressler). That seems to be the case based on a quick review of available images. </li></ul>



rd

Please Log in or Create an account to join the conversation.

Time to create page: 0.467 seconds
Powered by Kunena Forum