The difference between 'confidence_threshold' and 'match_conf'?

[Sherlock-IT]

what is the difference between 'confidence_threshold' and 'match_conf'.

[3nol]

The naming can be a bit confusing, and indeed, both describe some form of confidence. You can imagine confidence as a filter for when to keep or when to throw away something. This something is where these parameters are different:

• match_conf is passed to the feature matcher to describe which best match to keep for a given feature point. According to OpenCV's documentation, this parameter is used to check whether pairs of matches are dissimilar enough to be compared to each other:

  Features matcher, which finds two best matches for each feature and leaves the best one only if the ratio between descriptor distances is greater than the threshold match_conf.

  In the stitching package, the default value for ORB features is 0.3, for all other descriptors it is 0.65

• confidence_threshold describes how confident we are that two images are from the same panorama. The parameter is used to filter out non-fitting images for the panorama result. Also, it is used to parameterize the bundle adjuster. In the stitching package, the default value is 1.0.

[nndei]

Hello, thank you for the clarification. About confidence_threshold, I have seen this parameter called another way in all_relevant_matches = matcher.draw_matches_matrix, where the parameter is called conf_thresh. Meanwhile, the Subsetter class uses confidence_threshold. Is there a difference between these two?

Should one tweak the 0.3 match_conf parameter or the confidence_threshold parameter?

[lukasalexanderweber]

this is just a naming issue. You could rename conf_thresh to confidence_threshold

[nndei]

Thank you. So, in the tutorial code if I change one parameter I need to change the other one accordingly, otherwise things are a bit clashing? Meaning changing conf_thresh within matcher.draw_matches_matrix and also within Subsetter.

[lukasalexanderweber]

Yes, you need to change both :)