In this video, I explored the transition of Structure from Motion (SfM) from the traditional colmap pipeline to the more modern VGGSfM approach. I started by talking about Two-View Geometry (chapter 9 from Multiple View Geometry in Computer Vision by Hartley and Zisserman). Then I explored the implementations of colmap and VGGSfM from their GitHub repositories.
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
🔗 notes: https://mashaan14.github.io/YouTube-channel/nerf/2025_01_25_sfm
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
- website https://mashaan14.github.io/mashaan/
- github https://github.com/mashaan14
- X https://twitter.com/mashaan_14
- linkedin https://linkedin.com/in/mashaan
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
- sponsor this channel on GitHub Sponsors:
https://github.com/sponsors/mashaan14
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
📹 Video edit: DaVinci Resolve
🎧 Audio enhancement: Adobe Podcast
🖼️ Thumbnails: GIMP
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Chapters:
00:00 start
01:11 how SfM is related to NeRF?
03:24 epipolar geometry
09:41 fundamental matrix
10:33 essential matrix
11:52 colmap
14:13 initialization
20:24 image registration
21:00 triangulation
21:28 bundle adjustment
22:37 VGGSfM
24:54 tracker
29:09 camera initializer
31:48 triangulator
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
#StructureFromMotion #SfM #COLMAP #VGGSfM #NeRF #NeuralRadianceFields #TwoViewGeometry #EpipolarGeometry #ComputerVision #3DReconstruction #DeepLearningTutorial #MachineLearningProject #AIResearch #CodingTutorial