CODEWITHSUNDEEP
pythonopencvfeature-detectionorb
janedoe
janedoe

Reputation: 33

Bug? ORB returns more features than model allows

I want to compare feature vectors of different images for image recognition. The feature vectors are extracted with the following code:

orb = cv.ORB_create(nfeatures=500)
img = cv.imread(file, 0)
kp, des = orb.detectAndCompute(img, None)

The des-vector containing the descriptors/features should now be of shape (500x32) and it usually is. But in around 20% of the cases I get a feature vector like (501x32) or (502x32).

How can it happen that opencv returns more than nfeatures?

Upvotes: 1

Views: 79

Answers (0)

Related Questions