Why is face recognition sometimes inaccurate?

The two most common causes of inaccurate face recognition are:

  1. too few pixels per foot (PPF)
  2. non-optimal viewing position (e.g. profile angle vs. head-on)

Precise facial recognition requires a pixel resolution (Pixels Per Foot, aka PPF) that may not be available for all scenes that your security cameras monitor. For example, this is the full resolution image of the face detected in this scene:
bob.png

Even though this camera is HD, the face itself is only 92 PPF, because the average height of a male head is about 9 inches and the height of this cropped face is 69px. (i.e. 69 pixels / (9"/12") = 92 PPF)
bob_in_scene_Screen_Shot_2018-04-22_at_9.47.22_AM.png

When the total population of people to recognize is small, then it may be helpful for Camio to guess faces even when the PPF is too low to be precise. However, if you need faces to be recognized more precisely, then you can set a minimum PPF per face so that Camio avoids attempting to label faces detected at resolutions lower than your PPF minimum. The minimum PPF for face recognition can be set in the advanced options of Box-connected cameras on your camio.com/boxes page.

Learn more about how to search for a specific person.

Have more questions? Submit a request

Comments