Welcome, you are visitor number 
Overview
Visual Odometry is a subfield of computer vision where we aim to estimate the egomotion of an agent through camera images. In this project monocular images from a calibrated camera were used to estimate camera poses and the overall trajectory. The pipeline was implemented using different feature matching and tracking approaches for both 2D-2D and 3D-2D pose estimation techniques.
Implementations
Visual Odometry (VO) using 2 approaches for motion estimation
The first one is Structure From Motion(SFM) which is aimed to generate a sparse 3D reconstruction of a scene using random camera images
- 2D-2D - Feature Matching and Feature Tracking
- 3D-2D - Feature Matching and Feature Tracking
The first one is Structure From Motion(SFM) which is aimed to generate a sparse 3D reconstruction of a scene using random camera images
2D-2D Approach for Motion Estimation
3D-2D approach for motion estimation
|
|
|
Reprojection error in 3D-2D approach
Sparse 3D reconstruction.
|
|
Structure of Motion (SFM):
.
Team
- Arihanr Gaur
- Kushagra Srivastava
- Saurabh Kemekar
- Bipasha Parui
- Karthik Raman
- Parees Pathak
- Sushlok Shah
