Leveraging Camera Triplets for Efficient and Accurate Structure-from-Motion

Lalit Manam · Venu Madhav Govindu

Paper | Suppl | Project Page

This step is run after two-view-geometry calculation.

Typical pipeline:

1. Feature extraction
2. Image matching
3. Two-view-geometry calculation (Fundamental matrix calculation)
4. Camera Triplet scoring and view graph pruning
5. Scene reconstruction

This prorotype is implemented in a Jupyer notebook.

Requirements:

• networkx
• pycolmap
• tqdm

BibTeX

If you find our models useful, please consider citing the paper!

@inproceedings{manam2024camtripsfm,
  title                    = {Leveraging Camera Triplets for Efficient and Accurate Structure-from-Motion},
  author                   = {Manam, Lalit and Govindu, Venu Madhav},
  booktitle                = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  pages			   = {4959--4968},
  month			   = {June},
  year                     = {2024},
  organization             = {IEEE},
  Comment		   = {<a href="https://ee.iisc.ac.in/cvlab/research/camtripsfm/">[project page]</a>},
  url			   = {https://ee.iisc.ac.in/cvlab/research/camtripsfm/cam_triplets_sfm.pdf}
}