Another week, another blog about the latest features in 3.2 We’ve previously covered doorway analysis and this week it’s the turn of face recognition. This is a new analysis algorithm that, along with 3rd party face recognition servers, can find faces as you’re importing a video. Since this is a quite a large new feature we’ll be breaking it up over a few weeks. This week we’ll be looking at importing.
If you’re looking for specific people in a video then face recognition is your man. The face recognition import algorithm takes faces from the video and sends them to a 3rd party facial recognition server. This then sends back the resulting information to Kinesense which displayed it in both the Review tab and Grid tab (we’ll go into both of these in more detail in future posts).
When you go to import your video, select Face Recognition from the list of algorithms. The face detection settings will then appear below.
The face detect default settings usually do the trick but you may have to adjust them, especially if you are working with lower quality video.
Scale Factor: This parameter relates to how intensively the algorithm will search for faces in each frame. The lower number equals more intensive but slower processing.
Minimum Size: This parameter is the smallest expected size of a face in the video. This defaults to 1/16th of the frame height. If very small faces are expected (i.e. HD video with people in the distance) reduce this value. A smaller number will slow down processing. Minimum size is 40 pixels.
In the Advanced Settings section, there are more setting you can adjust.
Minimum Neighbours: This parameter relates detection sensitivity. Lower is more sensitive.
Processing Frame Frequency: This is the number of frames per second to be processed. The default of 10fps is generally suﬃcient.
IP Address of Face Server: This is the address of the computer that the 3rd party face recognition server is on. If you change this click Test Connection… to see if Kinesense can connect to the new address.
Once you’ve got all of you settings the way you want them to continue to import the video as normal.
Next week we’ll go into detail on how to review the imported video in the Review tab, view possible matches and add new matches. Until next time folk!
For more information, watch the video “Kinesense Face Recognition” on our YouTube Channel.