Better face detection

Better face detection

Face recognition is actually a combination of face detection and face recognition.   These two tasks work together.   

In this post we are discussing face detection as a separate step.  Refer to the post on "Better face suggestions" for a discussion of face recognition improvements planned.

If we continue to improve the accuracy of the face detector module, then that saves you time in two ways;
  1. avoiding having you delete "false positives" (not real faces)
  2. by not requiring that you add manual faces to correct "false negatives" (missed faces)
Factors in face detection include; resolution, lighting, pose, blur, occlusion.

We are testing a new machine learning face detector that is more accurate (>20% improvement).  The downside is that it is slower and takes up lots of memory... so we are trying to minimize the negative impact before we roll it into production.  

Tips and tricks: 
  1. our minimum face size is approx. 100 pixels in width 
  2. generally it is less effort (keyboard clicks) to delete false positives then it is to uncover and add false negatives.
  3. one trick to note - by doubling the image size (simply by blowing it up) so that the smallest faces are greater than 60 pixels in width that can improve the results in most cases if faces are missed on small pixel sized images (roughly a min width of 800 pixels is something to shoot for).  a good tool for that is XnConvert or XnResize - in the XnViewMP family.
    1. https://www.xnview.com/en/xnresize/ 
  4. we have embedded certain default settings for face detection confidence and quality
  5. if you have high resolution images then you can go with much higher confidence settings to avoid false positives
  6. by decreasing the confidence score more faces are picked up along with more errors; that is ok for low resolution images from email attachments and old scans of hardcopy photos; but not so good for higher resolution images.
  7. if you would like more information on these low level settings we can share them with you and how to "tune" them.  but we suggest that you only do this if you are technical and are confident editing configuration files.
Another tip is to enhance the quality of your images - if you are dealing with old scans.  One tool available is available at MyHeritage - example enhancement steps below.



    • Related Articles

    • Better face suggestions

      Face recognition is actually a combination of face detection and face recognition.   These two tasks work together.   Suggestions result from the face recognition step.    In this post we are discussing face recognition as a separate step.  Refer to ...
    • Add an "are you sure" request on the face thumbnail regeneration option

      Some users select many, many faces while tagging...using Shift-Click or Ctrl-Click. in the majority of cases to delete the faces.  But, if they select a large group of faces, right-click on one of the faces, and want to choose "Delete", but ...
    • Exporting faces

      We currently offer the ability to export photos selected via a complex search.  Those can be used for slideshows, books or other purposes. But, some users would like to export just the face thumbnails themselves.  So this would be feature added to ...
    • Enhancing images

      Do you have some poor resolution images?  Perhaps they came from scans of bad images or the photographer (we are not pointing fingers at you) wasn't very good on that day? There is an online service for enhancing photos that uses artificial ...
    • Better button layout to reduce mouse travel

      It was suggested from one of our users, Ralf, to better organize the popular button click sequences to reduce the need for mouse travel. We will also try to make more common tasks tied to a keyboard shortcut.  Remember that Enter will also mimick a ...