/Spherical-3D-Reconstruction-Pipeline

Primary LanguagePython

Spherical 3D Reconstruction Pipeline

PROPOSED METHOD OVERVIEW

  1. METHOD 1

Method 1 used OpenMVG to do camera pose calculation and then it uses OpenMVS for Densify the generated pointcloud.

  1. METHOD 2

Method 2 uses AliceVision to preprocesse 360 images and then it uses Colmap sfm tool to generate sparse and dense pointcloud file.

SETUP

1. DATA PREPARATION

FOLDER STRUCTURE

.
└── data/
    ├── config.yaml
    └── images/
        ├── image_1.jpeg
        ├── image_2.jpeg
        └── image_N.jpeg

config.yaml contains different configuration to run the proposed methods. For details check README.md of individual method.

RUN PIPELINE

Individual method folder has Dockerfile to run the 3D reconstruction pipeline.

To build the Docker image, run

cd <method dir>
docker build -t pipeline .

To run the 3D pipeline and generate .ply file, use the following command

docker run --gpus all -v <data folder path>:/app/data -t pipeline

It will save the final dense pointcloud .ply file.