Hi, thank you for releasing such a great work.
I observed that if I have a reference image of a single object, and a target image consisting of multiple instances of that object, LightGlue matches the keypoints of the reference image to the keypoints of exactly one instance of that object in the target image.
I'm wondering if there is any possibility of matching the object's keypoints to all the instances of the object in the target image.
A naive way I assume would be to treat it iteratively, i.e., take the first match, mask that object in the target image, run light glue again on the masked image to get the second match and so on.
Is there any possibility to get all the matches directly from LightGlue?
Hi, thank you for releasing such a great work.
I observed that if I have a reference image of a single object, and a target image consisting of multiple instances of that object, LightGlue matches the keypoints of the reference image to the keypoints of exactly one instance of that object in the target image.
I'm wondering if there is any possibility of matching the object's keypoints to all the instances of the object in the target image.
A naive way I assume would be to treat it iteratively, i.e., take the first match, mask that object in the target image, run light glue again on the masked image to get the second match and so on.
Is there any possibility to get all the matches directly from LightGlue?