Structure from motion exploration

With just two frames

feature matching

point cloud reconstruction is

With 13 frames it is

This is a very basic rudimentary SfM. For the same 13 input images Colmap reconstructs both sparse and dense in 2-3 minutes, which is incredible in quality and speed:

Sparse

Dense

This comes with

• Chopping frames from video file
• Camera calibration
• Reconstruction (from frame directory to point cloud PLY)