Using custom features and SIFT features

[apurva...@gmail.com]

Hello,

I have already added cameras, images and custom features to the COLMAP db. I want to  run the feature_extractor but it expects to add information like the "camera_model" and "camera_params" but I have already added this information in the db using database.py.

Any idea how I can bypass this to use feature extractor and matching?

[Wayne Cochran]

If you have already added the features to the db, why do you want to run the feature_extractor? Sounds like you are ready to perform matching.

[Apurva Zaveri]

Thanks much for the reply. The features I added are just a few pre-known features without any descriptors and I manually added matches for those in the database.

I do need SIFT features in addition to the ones I added. Hope that helps.

--
You received this message because you are subscribed to a topic in the Google Groups "COLMAP" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/colmap/MeJkCov-d1s/unsubscribe.
To unsubscribe from this group and all its topics, send an email to colmap+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/colmap/2f284e2f-dba1-44d1-8e5c-b02f6d5b91edn%40googlegroups.com.

[Wayne Cochran]

I assume when you run the feature_extractor on the same images it overwrites the custom features you already added? 
I don't know what happens when you run one of the feature matchers using features that lack a descriptor -- I would avoid that.

You could run the feature_extractor first and then run one of the matchers (e.g. exhaustive_matcher) . After both of those are finished then you

could add your custom features and matches to the DB. If you know the camera model and parameters you can add those to the DB at the very beginning
and make sure to specify --ImageReader.existing_camera_id for the feature_extractor. 

[apurva...@gmail.com]

Thank you for your response. Please see my response below:

On Monday, July 18, 2022 at 8:27:15 AM UTC-7 Wayne Cochran wrote:

I assume when you run the feature_extractor on the same images it overwrites the custom features you already added? 

Yes, that is correct.

 

I don't know what happens when you run one of the feature matchers using features that lack a descriptor -- I would avoid that.

COLMAP reconstructs just fine based on the custom keypoints and matches (without descriptors) that I inserted into the db manually.

You could run the feature_extractor first and then run one of the matchers (e.g. exhaustive_matcher) . After both of those are finished then you

could add your custom features and matches to the DB. If you know the camera model and parameters you can add those to the DB at the very beginning
and make sure to specify --ImageReader.existing_camera_id for the feature_extractor. 

Gotcha! Thanks, let me try that and report how it goes.