Open muhammadhemdan opened 6 years ago
Apologies for the late response.
Short Answer: Yes
So long as we can represent images as vectors where cosine similarity (or euclidean distance, or Jaccard distance) between the embedded vectors is a good indicator of semantic similarity, we can totally use it.
ORB, as we can say, is similar to SIFT and so we need to extract ORB and do some Bag-of-words embeddings (where distance relates to semantic similarity) and then we can use FLASH. We have done Images matching in past using similar algorithm (500x cheaper than openCV): Check out: https://www.cs.rice.edu/~as143/Papers/CAPSULE.pdf https://www.cs.rice.edu/~as143/presentations/CaPSuLe.pptx
as well as WTA (winner takes all) https://arxiv.org/abs/1612.01834
Anshumali
On Mon, Jan 15, 2018 at 6:55 AM, muhammadhemdan notifications@github.com wrote:
Hello, Can I use this algorithm to match two images? I'm using opencv ORB to detect and compute the key points and the descriptors.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/RUSH-LAB/Flash/issues/2, or mute the thread https://github.com/notifications/unsubscribe-auth/ANsGJLl0azB6R8EWvJv0p3cH0x_LtTPJks5tK0rdgaJpZM4ReX6T .
Hello, Can I use this algorithm to match two images? I'm using opencv ORB to detect and compute the key points and the descriptors.