cvg/pixel-perfect-sfm
A structure-from-motion and visual localization framework that refines keypoints, camera poses, and 3D points using deep feature alignment.
Not currently ranked — collecting fresh signals.
star history
This research project provides a Python package (pixsfm) that integrates with COLMAP and hloc for improving 3D reconstruction accuracy. It refines structure-from-motion results through two-step featuremetric optimization: first adjusting 2D keypoints jointly before SfM, then refining 3D points and camera poses via bundle adjustment. The refinement uses pre-trained CNN features extracted from images to minimize consistency across multiple views.
Frequently asked
- What is cvg/pixel-perfect-sfm?
- A structure-from-motion and visual localization framework that refines keypoints, camera poses, and 3D points using deep feature alignment.
- Is pixel-perfect-sfm open source?
- Yes — cvg/pixel-perfect-sfm is open source, released under the Apache-2.0 license.
- What language is pixel-perfect-sfm written in?
- cvg/pixel-perfect-sfm is primarily written in C++.
- How popular is pixel-perfect-sfm?
- cvg/pixel-perfect-sfm has 1.5k stars on GitHub.
- Where can I find pixel-perfect-sfm?
- cvg/pixel-perfect-sfm is on GitHub at https://github.com/cvg/pixel-perfect-sfm.