CSCI599

I performed this project as a part of the CSCI-599 3D Vision course I took in the Spring 2024 semester.

Table of Contents

• How to use
• Assignment 1: Geometry Processing
  • Introduction
  • Requirements / Rubric
• Assignment 2: Structure From Motion
  • Introduction
  • Requirements / Rubric

Introduction

In this assignment, I implemented structure from motion in computer vision. Structure from motion (SFM) is a technique used to reconstruct the 3D structure of a scene from a sequence of 2D images or video frames. It involves estimating the camera poses and the 3D positions of the scene points.

The goal of SFM is to recover the 3D structure of the scene and the camera motion from a set of 2D image correspondences. This can be achieved by solving a bundle adjustment problem, which involves minimizing the reprojection error between the observed 2D points and the projected 3D points.

To implement SFM, I performed the following steps:

1. Feature extraction: Extract distinctive features from the input images.
2. Feature matching: Match the features across different images to establish correspondences.
3. Camera pose estimation: Estimate the camera poses for each image.
4. Triangulation: Compute the 3D positions of the scene points using the camera poses and the corresponding image points.
5. Bundle adjustment: Refine the camera poses and the 3D points to minimize the reprojection error.

To exexcute the program:

python ./assignments/assignment2/feat_match.py --data_dir "./assets/DATASET-NAME/images/" --out_dir "./assets/DATASET-NAME/"
python ./assignments/assignment2/sfm.py --data_dir "./assets/" --dataset "DATASET-NAME"