Skip to content

Conversation

mihaidusmanu
Copy link
Contributor

@mihaidusmanu mihaidusmanu commented Jun 9, 2021

I implemented a simple interface for VLFeat SIFT from PyCOLMAP.

There might some issue with top keypoints selection - I opened an issue on the PyCOLMAP repository to follow up colmap/pycolmap#8.

I get the following results on South-Building without image downsampling and max_keypoints=10000 (to avoid GPU memory issues during matching):

{
  'mean_reproj_error': 0.450604,
  'mean_track_length': 7.001359,
  'num_input_images': 128,
  'num_observations': 535786,
  'num_observations_per_image': 4185.828125,
  'num_reg_images': 128,
  'num_sparse_points': 76526
}

which seems reasonable 😄.

sarlinpe pushed a commit that referenced this pull request Jun 26, 2021
* Added SIFT features using PyCOLMAP.

* Added asserts for input image.

* Fix style

Co-authored-by: Mihai Dusmanu <mihai.dusmanu@gmail.com>
See #69
@sarlinpe
Copy link
Member

Merged in #72.

@sarlinpe sarlinpe closed this Jun 26, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants