Cut the photo without losing faces

asked 2014-05-05 16:16:45 -0500

Sergey gravatar image

Hello. Can you help me ? I am doing a small iOS-application and I want manually parallelize face detection (CascadeClassifier) using Grand Central Dispatch. For example, take the photo, divide it into two parts and run face detection for two parts of photo in different threads. Question - how to cut the photo? Suddenly it can be cutted in the face, and then it is not defined in any of the threads. Is there any way to cut the photo so that is guaranteed not to damage the face? But do not run a search of all faces in the photo?

What I came up with: 1) Before cutting the photo - search for the face in the central part of the photo (not the entire photo), it’s not good. 2) try to find the eyes in the photo - but it is too long.

edit retag flag offensive close merge delete



Why do you need to parallelize face detection? isn't it fast enough?

GilLevi gravatar imageGilLevi ( 2014-05-06 11:41:43 -0500 )edit

GilLevi, Yes, it's fast, but I want even faster. I need maximum performance for live detection (and recognition then).

Sergey gravatar imageSergey ( 2014-05-06 12:39:37 -0500 )edit