A way to manually flag matches as wrong that can help Colmap rematch an image pair ignoring the previously paired features and try matching them somewhere else?

[jared-krauss]

For instance, a van that has a sign on both sides that is exactly the same size shape lettering color etc so many features to match, but it misplaces where the photo is viewing from because of that. But if there were a way to, when viewing the matched features in the database GUI to indicate which pairs should be retried (or in the CLI, but I can't visualize that), then retry the feature matching on that image pair but with the knowledge that the way it tried to match those set of features last time were wrong.

Because with masking, you can sometimes take away very good identifiable features when trying to prevent mismatching like this.

Or I have to manually break down all the image pairs into small incomplete models, then merge them together later, but that's so tedious of an idea I haven't tried implementing it, but I am slowly tracking which sets of images work best together etc to have a headstart when I get there.

I'm using my own dataset, a city night scene.

[jared-krauss]

Sorry ya'll from when I very first got on to github and didn't know where i was posting and thought I was posting int he Colmap discussion ahah