OpenCV-Python Dense SIFT Settings

Question

This is a follow-up question to the previously posted question about using OpenCVs dense sift implementation in python (OpenCV-Python dense SIFT).

Using the suggested code for a dense sift

    dense=cv2.FeatureDetector_create("Dense")
    kp=dense.detect(imgGray)
    kp,des=sift.compute(imgGray,kp)

I have the following questions:

Can I access any of the DenseFeatureDetector properties in python? Set or at least read?
What is the logic behind c++s FeatureDetector::create becoming pythons FeatureDetector_create? How can I know that based on the documentation (http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_feature_detectors.html)?
Any recommendation on a python wrapper for VLFeat Library? Is pyvlfeat still the way to go (I tried to setup pyvlfeat but it didn't compile on my mac)?

Thanks!

maybe you can just query the builtin docs: `dense=cv2.FeatureDetector_create("Dense"); help(dense)` also be aware, that the interfaces for opencv3.0 have changed dramatically here, and won't allow you to use dense anymore, so careful. — berak, Apr 30 '15 at 14:23
I see; there will be no dense sift at all anymore or only not for python? — Legie, Apr 30 '15 at 15:04
no, in general. but if you think of it, you can easily create your own keypoints — berak, Apr 30 '15 at 15:19

score 3 · Answer 1 · answered Aug 27 '15 at 10:29

You can see current (default) options with the following:

dense = cv2.FeatureDetector_create('Dense')
f = '{} ({}): {}'
for param in dense.getParams():
    type_ = dense.paramType(param)
    if type_ == cv2.PARAM_BOOLEAN:
        print f.format(param, 'boolean', dense.getBool(param))
    elif type_ == cv2.PARAM_INT:
        print f.format(param, 'int', dense.getInt(param))
    elif type_ == cv2.PARAM_REAL:
        print f.format(param, 'real', dense.getDouble(param))
    else:
        print param

Then you'd get an output like the following:

featureScaleLevels (int): 1
featureScaleMul (real): 0.10000000149
initFeatureScale (real): 1.0
initImgBound (int): 0
initXyStep (int): 6
varyImgBoundWithScale (boolean): False
varyXyStepWithScale (boolean): True

You can alter the options as the following:

dense.setDouble('initFeatureScale', 10)
dense.setInt('initXyStep', 3)

OpenCV-Python Dense SIFT Settings

1 Answers1

Linked