3

I am trying to process faces in images on a device/tablet. I have used opencv (NDK) a while ago. I see that there are couple of other options available to process faces. Just wondering how opencv, android-vision api and FastCV would compare specifically for processing faces;

Found couple of similar posts here but they did not answer all my questions.

Android API face detection vs. OpenCV/JavaCV face detection

Android Computer Vision JavaCv OpenCV Fastv comparison

My questions: 1) How does android face api (face detection and landmark extraction) compare to opencv (javaCV or opencv NDK) (accuracy Vs speed) 2) Is FastCV better for this? I presume it comes with license restrictions. 3) Does the android api work for all android devices?

I found couple of ports for open-source CV libraries (other than opencv) that are commonly used for pc based CV applications but these are not optimized for the devices, I found them very slow when used as is. thanks!

Community
  • 1
  • 1
eyebies
  • 39
  • 6
  • What did you end up using? I am using android vision api but if detector's All classification is enabled it slows down performance. – hardik9850 Jun 15 '16 at 09:43
  • 1
    I had to take the traditional path (javacv) for non-technical reasons. I was more interested in the key points in the vision api as I could use them for alignment and registration. I know this depends on the phone we use for testing but how much time does it usually take for face detection and the fiducial poiny extraction? I think the relative timing info should be comparable across different phones. – eyebies Jun 16 '16 at 13:52

0 Answers0