While I am training the SuperGlue model, I am using SIFT detector based keypoints [from Github repo]. But the scores values in it appears to be very less (like somewhere around less than 0.1). When I am checking the scores of the keypoints using Superpoint it appears to be around 0.1 to 0.9.
On using SIFT detector, I am changing the match_threshold value in models/superglue.py to be 1.0 (from 0.2). That generates the image pairs but no matches at all. Can someone tell me what am I doing wrong here?