Visual Odometry pipeline

Monocular, visual odometry (VO) pipeline with the most essential features:

1. initialization of 3D landmarks
2. keypoint tracking between two frames
3. pose estimation using established 2D↔3D correspondences
4. triangulation of new landmarks.

Results

Here is a quick overview of our results: https://www.dropbox.com/sh/jvh6bk42ok4fu2e/AADP0zjf7AOGADLhL-1Z81gna?dl=0

Installation

1. To install this source code, first clone the repo:

git clone [email protected]:ToniRV/visual-odometry-pipeline.git

2. Modify the paths to the datasets that you can download from here Copy this file and rename it to paths.txt Modify your paths inside this file accordingly. Then store it in the Parameters folder.

3. Now you can head to matlab and open main. You can now press Run.

Reference output

https://www.youtube.com/watch?v=MQ1iDi3S-TU&list=PLI5XgAFFHqZiAg6yd0XKOTRgkCE4QTm54&index=3#t=32.749926