Monday, 11 February, 2013 § Leave a comment
A while back i was working on a interesting project for an Autonomous Realtime Image Analytics Engine a component of this engine was fast face detection and tracking using existing classifiers available in OpenCV but improving their performance to get the best result possible.
I was able to get a stable 50fps for face detection and tracking for up 10 faces.
- Some optimizations that were made were to achieve this were
- Threading of the entire operation
- skin segmentation
- defining ROIs
- and contour detection