This repository contains the following functionality/algorithm implementations for SfM:
| Functionality | Implementation | Class Name |
|---|---|---|
| feature detection | OpenCV | view_processor (VP) |
| feature matching | OpenCV | key_tracker (KT) |
| 2D and 3D point tracking | Self | key_tracker |
| fundamental matrix calculation | Self | epipolar_processor (EP) |
| essential matrix calculation | Self | epipolar_processor |
| camera pose extraction from essential matrix | Self | campose_processor (CP) |
| camera pose disambiguation | Self | campose_processor |
| linear triangulation | Self | triangulation_processor (TP) |
| nonlinear triangulation | Self | triangulation_processor |
| linear perspective-n-point estimation | Self | campose_processor |
| nonlinear perspective-n-point estimation | Self | campose_processor |
| bundle adjustment | Self | ba_processor (BAP) |
The pipeline (data flow) is as shown below
- Python 3.8.17
- Conda environment: environment_sfm.yml
Each script has its own unit. Ex. python3 epipolar_processor.py. For the whole pipeline testing, python3 ba_processor.py.
- https://www.cis.upenn.edu/~cis580/Spring2015/Projects/proj2/proj2.pdf
- https://www-users.cse.umn.edu/~hspark/CSci5980/hw5.pdf
- https://www-users.cse.umn.edu/~hspark/csci5563_S2021/hw4.pdf
- https://cvgl.stanford.edu/teaching/cs231a_winter1415/lecture/lecture3_camera_calibration_notes.pdfhttps://cvgl.stanford.edu/teaching/cs231a_winter1415/lecture/