I'm trying to implement a system that is able to predict the type of scene between 8 cases through images. To do that, I'm implementing the Bag of Visual Words algorithm with different local descriptors and classifiers. I'm testing SIFT and KAZE and its Dense versions (instead of using the keypoints found by the descriptors, I've created a mesh of keypoints all over the image:
Keypoints of the different descriptors
However, when I'm trying to compute the KAZE features from the DENSE keypoints:
# Create KAZE object
KAZE_detector = cv2.KAZE_create(threshold=0.0001)
# Parameter to adjust the size of the keypoints
step_size = 8
# Read image and transform to gray level
ima=cv2.imread(train_images_filenames[0])
gray=cv2.cvtColor(ima,cv2.COLOR_BGR2GRAY)
# Create the DENSE keypoints
kpt = [cv2.KeyPoint(x, y, step_size) for y in range(0, gray.shape[0], step_size) for x in range(0, gray.shape[1], step_size)]
# Compute the features
kpt, des = KAZE_detector.compute(gray, kpt)
The following error occurs:
---------------------------------------------------------------------------
error Traceback (most recent call last)
/tmp/ipykernel_4330/2348194759.py in <module>
13
14 # Compute the features
---> 15 kpt, des = KAZE_detector.compute(gray, kpt)
16
error: OpenCV(4.5.3) /tmp/pip-req-build-afu9cjzs/opencv/modules/features2d/src/kaze/KAZEFeatures.cpp:554: error: (-215:Assertion failed) 0 <= kpts[i].class_id && kpts[i].class_id < static_cast<int>(evolution_.size()) in function 'Feature_Description'
I do not understand why this happens because I use the same exact piece of code for DENSE SIFT and it is working properly.
