alvaro-budria/3D-Recovery-of-Urban-Scenes

We perform 3D reconstruction of buildings from sets of images of urban scenes taken from different points of view.

3D Reconstruction of Urban Scenes

Overview

This repository contains the code for the "3D Recovery of Urban Scenes" project from the Master in Computer Vision at UAB (Module 4: 3D Vision).

The project focuses on reconstructing 3D models of buildings using multiple images taken from different perspectives, including frontal façade images and aerial images. The tasks involve rectifying perspective distortion, estimating essential and fundamental matrices, calibrating cameras with planar patterns, estimating depth, generating new views, and creating 3D models using calibrated or uncalibrated cameras (SfM).

To understand the complete pipeline, refer to the specific code directory and technical report for each week's delivery (listed below).

Deliverables

The project is divided into five different weekly deliveries, each contributing to the overall pipeline. Below are the links to the respective weeks' code and explanations:

Week 1: Planar Transformations and Image Rectification

• Code

• Report

Week 2: Homographies for Panorama Reconstruction, Camera Calibration, and Logo Replacement

• Code

• Report

Week 3: Epipolar Geometry of Two Views and Photo-Sequencing

• Code

• Report

Week 4: Reconstruction from Two Images with Known Internal Parameters

• Code

• Report

Week 5: Structure from Motion

• Code

• Report

Final Presentation

The final presentation, which explains our approach in detail, can be accessed here.

Contributors: Team 7

The project was completed by the following contributors:

• Adria Molina
• Sergi Masip
• Alex Carrillo
• Alvaro Budria