/3DReconstruct

Object Mesh Reconstruction using Azure Kinect

Primary LanguageJupyter Notebook

Object Mesh Reconstruction using Azure Kinect

The whole framework is the post/preprocess of BADSLAM (https://github.com/ETH3D/badslam) mapping result.

Point Cloud generation

See README.md at pointCloudGen/README.md

Background removal

Source code: backgroundRemoval/backgroundremoval.ipynb

Dependencies: open3d library

open3d install:

pip install open3d

Geometric registration

Source code: geometricRegistration/ransac_icp_colored_regi.py

Dependencies: open3d library

Surface Reconstruction

Source code: MeshRecon/Mesh_recon.py

Dependencies: open3d library

Depth error compensation

Training data generation source code: biasCorrection/april_generate_traiing_data.ipynb

Learning source code:

--Random Forest: biasCorrection/training_RF.ipynb

--Neural Network: biasCorrection/training.ipynb

Dependencies: open3d, AprilTag (https://github.com/AprilRobotics/apriltag), tensorflow, sklearn

Setup AprilTag:

$ git clone https://github.com/AprilRobotics/apriltag.git
$ cd apriltag
$ cmake .
$ sudo make install

Setup other libraries:

pip install open3d
pip install tensorflow==2.0.0-rc1
pip install sklearn