YOLO v2 anchors generated method _K-means clustering

YOLO v2 anchors

in .cfg file

[region]
anchors =  1.3221, 1.73145, 3.19275, 4.00944, 5.05587, 8.09892, 9.47112, 4.84053, 11.2364, 10.0071    # height,width each anchor
bias_match=1
classes=4
coords=4
num=5
softmax=1
jitter=.3
rescore=1

instead of choosing priors by hand, we run k-means clustering on the training set bounding boxes to automatically find good priors.

k-means algorithm

randomly initialize k cluster centroids u1,u2,...,uk

repeat{

   # step 1 : cluster assignment for training data

   for i =1 to m

        c(i):=index(from 1 to k) of cluster centroid

                 closest to x(i)

   # step 2 : move centroid

   for k=1 to K

        uk :=average(mean) of points assigned to cluster k

}