The face detector i made, works in neural network basis. The main idea is to search all pixel in images to find facial area by using keyblock. But before that, i trained the neural network so that it would be able to distinguish facial and non-facial area.
Here is some of the detection result from my face-detector:

The result is in RGB mode, while the detection process on gray-scale mode. As you can see, the facial image shown a face size variation. And my face detector also able to detect face and non face images, like dogs.
Next the algorithm :
1. make sliding windows to be looped on every pixel in the image (i use 30x30 sliding windows size)
2. once the loop finished, re-size the image to be smaller with a pre-determined scale-down ratio (i use 0.8)
3. stop when the size of image is smaller than the size of sliding windows.
4. eliminate false detection
5. extract final position result
No comments:
Post a Comment