Opencv-python. How can I use convertTo?

Python
1 
brisk = cv2.BRISK_create()
des = brisk.compute(img,kp, None)
des.convertTo(des_left,cv2.CV_32F)

> AttributeError: ‘tuple’ object has no attribute ‘convertTo’

or

des = des[1].convertTo(des,cv2.CV_32F)

> AttributeError: ‘numpy.ndarray’ object has no attribute ‘convertTo’

How can I use convertTo in opencv-python? Sincerely hope to get your help.

2

use numpy astype

3

please explain, WHY you try to do this.

there is no meaningful conversion to float from binary descriptors (BRISK,ORB,etc), those are bitstrings , not numbers

again, please explain the context, seems you try to blindly copypaste a broken idea :wink:

4

I want to try fast detector+brisk descriptor+flann matching, and I found the brisk descriptor don’t fit the input type. After googling , I found someone recommend to use convertTo to make it possible. I tried, but convertTo seems to not included in the opencv-python(contrib), so I came here to ask :rofl:.

By the way, if I achieve the convertTo, do the matching outcomes meaningful? Or bfmatch will be a better choice?

5

no way. the result is just garbage.

keep the descriptors as-is (uint8) and match them with BFMatcher (HAMMING), or . IF you need to use flann, use an LSHIndex:

FLANN_INDEX_LSH = 6

index_params= dict(algorithm = FLANN_INDEX_LSH,
                   table_number = 6, # 12
                   key_size = 12,     # 20
                   multi_probe_level = 1) #2
search_params = dict(checks=50)   # or pass empty dictionary

flann = cv2.FlannBasedMatcher(index_params,search_params)