about match_keypoints

[cxb1998]

Can you provide a demo for the use of custom point matching?
thanks

[cxb1998]

Or give some detailed notes

[tianjiahao]

I get result in demo_fundamental.py, kpts1, kpts2:

[tianjiahao]

[cxb1998]

I get result in demo_fundamental.py, kpts1, kpts2:

hello! Your image seems to be damaged, can you upload it again? Also, are the results you get referring to the custom points?

[tianjiahao]

I get result in demo_fundamental.py, kpts1, kpts2:

hello! Your image seems to be damaged, can you upload it again? Also, are the results you get referring to the custom points?

I add this code to demo_fundamental.py, you can try.
img1 = cv2.imread(im1_path)
img2 = cv2.imread(im2_path)
for i,v in enumerate(mask):
if v[0]==1:
m1 = kpts1[i].cpu().numpy().astype('int32')
m2 = kpts2[i].cpu().numpy().astype('int32')
cv2.circle(img1, m1, 5, (0,0,255), -1)
cv2.circle(img2, m2, 5, (0,0,255), -1)
cv2.imwrite('/data/tjh/RoMa/output/sacre-1.png', img1)
cv2.imwrite('/data/tjh/RoMa/output/sacre-2.png', img2)

[mafeng199325]

I am also eager to know how to generate such images using match_keypoints？

[Parskatt]

Hi, Im away for the summer, will get back with some demos start of september.

Regarding pic above, use the match_keypoint method for your keypoints of choice and use e.g. opencv to visualize.

[mafeng199325]

你好，我今年夏天不在家，九月初会带着一些演示回来。
Special look forward to, excellent algorithm let people worship. I hope to see a detailed explanation then, thank you

[mafeng199325]

I found that the algorithm did not work well for small objects and was not rotation invariant. Is there any solution?