probreg


Nameprobreg JSON
Version 0.3.8 PyPI version JSON
download
home_pagehttps://github.com/neka-nat/probreg
SummaryProbablistic point cloud resitration algorithms
upload_time2024-05-11 06:42:17
maintainerNone
docs_urlNone
authorneka-nat
requires_pythonNone
licenseNone
keywords
VCS
bugtrack_url
requirements No requirements were recorded.
Travis-CI No Travis.
coveralls test coverage No coveralls. 
            # ![logo](https://raw.githubusercontent.com/neka-nat/probreg/master/images/logo.png)
[![Build status](https://github.com/neka-nat/probreg/actions/workflows/build-and-test.yaml/badge.svg)](https://github.com/neka-nat/probreg/actions/workflows/build-and-test.yaml/badge.svg)
[![PyPI version](https://badge.fury.io/py/probreg.svg)](https://badge.fury.io/py/probreg)
[![MIT License](http://img.shields.io/badge/license-MIT-blue.svg?style=flat)](LICENSE)
[![Documentation Status](https://readthedocs.org/projects/probreg/badge/?version=latest)](https://probreg.readthedocs.io/en/latest/?badge=latest)
[![Downloads](https://static.pepy.tech/badge/probreg)](https://pepy.tech/project/probreg)

Probreg is a library that implements point cloud **reg**istration algorithms with **prob**ablistic model.

The point set registration algorithms using stochastic model are more robust than ICP(Iterative Closest Point).
This package implements several algorithms using stochastic models and provides a simple interface with [Open3D](http://www.open3d.org/).

## Core features

* Open3D interface
* Rigid and non-rigid transformation

## Algorithms

* Maximum likelihood when the target or source point cloud is observation data
    * [Coherent Point Drift (2010)](https://arxiv.org/pdf/0905.2635.pdf)
    * [Extended Coherent Point Drift (2016)](https://ieeexplore.ieee.org/abstract/document/7477719) (add correspondence priors to CPD)
    * [Color Coherent Point Drift (2018)](https://arxiv.org/pdf/1802.01516)
    * [FilterReg (CVPR2019)](https://arxiv.org/pdf/1811.10136.pdf)
* Variational Bayesian inference
    * [Bayesian Coherent Point Drift (2020)](https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8985307)
* Distance minimization of two probabilistic distributions
    * [GMMReg (2011)](https://ieeexplore.ieee.org/document/5674050)
    * [Support Vector Registration (2015)](https://arxiv.org/pdf/1511.04240.pdf)
* Hierarchical Stocastic model
    * [GMMTree (ECCV2018)](https://arxiv.org/pdf/1807.02587.pdf)

### Transformations

| type | CPD | SVR, GMMReg | GMMTree | FilterReg | BCPD (experimental) |
|------|-----|-------------|---------|-----------|---------------------|
|Rigid | **Scale + 6D pose** | **6D pose** | **6D pose** | **6D pose** </br> (Point-to-point,</br> Point-to-plane,</br> FPFH-based)| - |
|NonRigid | **Affine**, **MCT** | **TPS** | - | **Deformable Kinematic** </br> (experimental) | **Combined model** </br> (Rigid + Scale + NonRigid-term)

### CUDA support
You need to install cupy.

```
pip install cupy
```

* [Rigid CPD](https://github.com/neka-nat/probreg/blob/master/examples/cpd_rigid_cuda.py)
* [Affine CPD](https://github.com/neka-nat/probreg/blob/master/examples/cpd_affine3d_cuda.py)

## Installation

You can install probreg using `pip`.

```
pip install probreg
```

Or install probreg from source.

```
git clone https://github.com/neka-nat/probreg.git --recursive
cd probreg
pip install -e .
```

## Getting Started

This is a sample code that reads a PCD file and calls CPD registration.
You can easily execute registrations from Open3D point cloud object and draw the results.

```py
import copy
import numpy as np
import open3d as o3
from probreg import cpd

# load source and target point cloud
source = o3.io.read_point_cloud('bunny.pcd')
source.remove_non_finite_points()
target = copy.deepcopy(source)
# transform target point cloud
th = np.deg2rad(30.0)
target.transform(np.array([[np.cos(th), -np.sin(th), 0.0, 0.0],
                           [np.sin(th), np.cos(th), 0.0, 0.0],
                           [0.0, 0.0, 1.0, 0.0],
                           [0.0, 0.0, 0.0, 1.0]]))
source = source.voxel_down_sample(voxel_size=0.005)
target = target.voxel_down_sample(voxel_size=0.005)

# compute cpd registration
tf_param, _, _ = cpd.registration_cpd(source, target)
result = copy.deepcopy(source)
result.points = tf_param.transform(result.points)

# draw result
source.paint_uniform_color([1, 0, 0])
target.paint_uniform_color([0, 1, 0])
result.paint_uniform_color([0, 0, 1])
o3.visualization.draw_geometries([source, target, result])
```

## Resources

* [Documentation](https://probreg.readthedocs.io/en/latest/?badge=latest)

## Results

### Compare algorithms

| CPD | SVR | GMMTree | FilterReg |
|-----|-----|---------|-----------|
| <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/cpd_rigid.gif" width="640"> |  <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/svr_rigid.gif" width="640"> | <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/gmmtree_rigid.gif" width="640"> | <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_rigid.gif" width="640"> |

### Noise test

| ICP(Open3D) | CPD | FilterReg |
|-------------|-----|-----------|
| <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/icp_noise.gif" width="640"> |  <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/cpd_noise.gif" width="640"> | <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_noise.gif" width="640"> |

### Non rigid registration

| CPD | SVR | Filterreg | BCPD |
|-----|-----|-----------|------|
| <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/cpd_nonrigid.gif" width="640"> | <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/svr_nonrigid.gif" width="640"> | <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_deformable.gif" width="640"> | <img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/bcpd_nonrigid.gif" width="640"> |

### Feature based registration

| FPFH FilterReg |
|----------------|
|<img src="https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_fpfh.gif" width="640"> |

### Time measurement

Execute an example script for measuring time.

```
OMP_NUM_THREADS=1 python time_measurement.py

# Results [s]
# ICP(Open3D):  0.0014092829951550812
# CPD:  0.038112225010991096
# SVR:  0.036476270004641265
# GMMTree:  0.10535842599347234
# FilterReg:  0.005098833993542939
```

## Citing

```
@software{probreg,
    author = {{Kenta-Tanaka et al.}},
    title = {probreg},
    url = {https://probreg.readthedocs.io/en/latest/},
    version = {0.1.6},
    date = {2019-9-29},
}
```

Raw data

            {
    "_id": null,
    "home_page": "https://github.com/neka-nat/probreg",
    "name": "probreg",
    "maintainer": null,
    "docs_url": null,
    "requires_python": null,
    "maintainer_email": null,
    "keywords": null,
    "author": "neka-nat",
    "author_email": "nekanat.stock@gmail.com",
    "download_url": "https://files.pythonhosted.org/packages/ed/f5/4cda3ac2e064013c293f7b405a9ed2e55f80ef802f2e7811eb6e9ef6b094/probreg-0.3.8.tar.gz",
    "platform": null,
    "description": "# ![logo](https://raw.githubusercontent.com/neka-nat/probreg/master/images/logo.png)\n[![Build status](https://github.com/neka-nat/probreg/actions/workflows/build-and-test.yaml/badge.svg)](https://github.com/neka-nat/probreg/actions/workflows/build-and-test.yaml/badge.svg)\n[![PyPI version](https://badge.fury.io/py/probreg.svg)](https://badge.fury.io/py/probreg)\n[![MIT License](http://img.shields.io/badge/license-MIT-blue.svg?style=flat)](LICENSE)\n[![Documentation Status](https://readthedocs.org/projects/probreg/badge/?version=latest)](https://probreg.readthedocs.io/en/latest/?badge=latest)\n[![Downloads](https://static.pepy.tech/badge/probreg)](https://pepy.tech/project/probreg)\n\nProbreg is a library that implements point cloud **reg**istration algorithms with **prob**ablistic model.\n\nThe point set registration algorithms using stochastic model are more robust than ICP(Iterative Closest Point).\nThis package implements several algorithms using stochastic models and provides a simple interface with [Open3D](http://www.open3d.org/).\n\n## Core features\n\n* Open3D interface\n* Rigid and non-rigid transformation\n\n## Algorithms\n\n* Maximum likelihood when the target or source point cloud is observation data\n    * [Coherent Point Drift (2010)](https://arxiv.org/pdf/0905.2635.pdf)\n    * [Extended Coherent Point Drift (2016)](https://ieeexplore.ieee.org/abstract/document/7477719) (add correspondence priors to CPD)\n    * [Color Coherent Point Drift (2018)](https://arxiv.org/pdf/1802.01516)\n    * [FilterReg (CVPR2019)](https://arxiv.org/pdf/1811.10136.pdf)\n* Variational Bayesian inference\n    * [Bayesian Coherent Point Drift (2020)](https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8985307)\n* Distance minimization of two probabilistic distributions\n    * [GMMReg (2011)](https://ieeexplore.ieee.org/document/5674050)\n    * [Support Vector Registration (2015)](https://arxiv.org/pdf/1511.04240.pdf)\n* Hierarchical Stocastic model\n    * [GMMTree (ECCV2018)](https://arxiv.org/pdf/1807.02587.pdf)\n\n### Transformations\n\n| type | CPD | SVR, GMMReg | GMMTree | FilterReg | BCPD (experimental) |\n|------|-----|-------------|---------|-----------|---------------------|\n|Rigid | **Scale + 6D pose** | **6D pose** | **6D pose** | **6D pose** </br> (Point-to-point,</br> Point-to-plane,</br> FPFH-based)| - |\n|NonRigid | **Affine**, **MCT** | **TPS** | - | **Deformable Kinematic** </br> (experimental) | **Combined model** </br> (Rigid + Scale + NonRigid-term)\n\n### CUDA support\nYou need to install cupy.\n\n```\npip install cupy\n```\n\n* [Rigid CPD](https://github.com/neka-nat/probreg/blob/master/examples/cpd_rigid_cuda.py)\n* [Affine CPD](https://github.com/neka-nat/probreg/blob/master/examples/cpd_affine3d_cuda.py)\n\n## Installation\n\nYou can install probreg using `pip`.\n\n```\npip install probreg\n```\n\nOr install probreg from source.\n\n```\ngit clone https://github.com/neka-nat/probreg.git --recursive\ncd probreg\npip install -e .\n```\n\n## Getting Started\n\nThis is a sample code that reads a PCD file and calls CPD registration.\nYou can easily execute registrations from Open3D point cloud object and draw the results.\n\n```py\nimport copy\nimport numpy as np\nimport open3d as o3\nfrom probreg import cpd\n\n# load source and target point cloud\nsource = o3.io.read_point_cloud('bunny.pcd')\nsource.remove_non_finite_points()\ntarget = copy.deepcopy(source)\n# transform target point cloud\nth = np.deg2rad(30.0)\ntarget.transform(np.array([[np.cos(th), -np.sin(th), 0.0, 0.0],\n                           [np.sin(th), np.cos(th), 0.0, 0.0],\n                           [0.0, 0.0, 1.0, 0.0],\n                           [0.0, 0.0, 0.0, 1.0]]))\nsource = source.voxel_down_sample(voxel_size=0.005)\ntarget = target.voxel_down_sample(voxel_size=0.005)\n\n# compute cpd registration\ntf_param, _, _ = cpd.registration_cpd(source, target)\nresult = copy.deepcopy(source)\nresult.points = tf_param.transform(result.points)\n\n# draw result\nsource.paint_uniform_color([1, 0, 0])\ntarget.paint_uniform_color([0, 1, 0])\nresult.paint_uniform_color([0, 0, 1])\no3.visualization.draw_geometries([source, target, result])\n```\n\n## Resources\n\n* [Documentation](https://probreg.readthedocs.io/en/latest/?badge=latest)\n\n## Results\n\n### Compare algorithms\n\n| CPD | SVR | GMMTree | FilterReg |\n|-----|-----|---------|-----------|\n| <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/cpd_rigid.gif\" width=\"640\"> |  <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/svr_rigid.gif\" width=\"640\"> | <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/gmmtree_rigid.gif\" width=\"640\"> | <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_rigid.gif\" width=\"640\"> |\n\n### Noise test\n\n| ICP(Open3D) | CPD | FilterReg |\n|-------------|-----|-----------|\n| <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/icp_noise.gif\" width=\"640\"> |  <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/cpd_noise.gif\" width=\"640\"> | <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_noise.gif\" width=\"640\"> |\n\n### Non rigid registration\n\n| CPD | SVR | Filterreg | BCPD |\n|-----|-----|-----------|------|\n| <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/cpd_nonrigid.gif\" width=\"640\"> | <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/svr_nonrigid.gif\" width=\"640\"> | <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_deformable.gif\" width=\"640\"> | <img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/bcpd_nonrigid.gif\" width=\"640\"> |\n\n### Feature based registration\n\n| FPFH FilterReg |\n|----------------|\n|<img src=\"https://raw.githubusercontent.com/neka-nat/probreg/master/images/filterreg_fpfh.gif\" width=\"640\"> |\n\n### Time measurement\n\nExecute an example script for measuring time.\n\n```\nOMP_NUM_THREADS=1 python time_measurement.py\n\n# Results [s]\n# ICP(Open3D):  0.0014092829951550812\n# CPD:  0.038112225010991096\n# SVR:  0.036476270004641265\n# GMMTree:  0.10535842599347234\n# FilterReg:  0.005098833993542939\n```\n\n## Citing\n\n```\n@software{probreg,\n    author = {{Kenta-Tanaka et al.}},\n    title = {probreg},\n    url = {https://probreg.readthedocs.io/en/latest/},\n    version = {0.1.6},\n    date = {2019-9-29},\n}\n```",
    "bugtrack_url": null,
    "license": null,
    "summary": "Probablistic point cloud resitration algorithms",
    "version": "0.3.8",
    "project_urls": {
        "Homepage": "https://github.com/neka-nat/probreg"
    },
    "split_keywords": [],
    "urls": [
        {
            "comment_text": "",
            "digests": {
                "blake2b_256": "edf54cda3ac2e064013c293f7b405a9ed2e55f80ef802f2e7811eb6e9ef6b094",
                "md5": "8877127558e279094fffa87ac6d0c269",
                "sha256": "35558e43ddc68db999692818702c688a53fffb5168d535872edba890e9c6b5f8"
            },
            "downloads": -1,
            "filename": "probreg-0.3.8.tar.gz",
            "has_sig": false,
            "md5_digest": "8877127558e279094fffa87ac6d0c269",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": null,
            "size": 1040035,
            "upload_time": "2024-05-11T06:42:17",
            "upload_time_iso_8601": "2024-05-11T06:42:17.043478Z",
            "url": "https://files.pythonhosted.org/packages/ed/f5/4cda3ac2e064013c293f7b405a9ed2e55f80ef802f2e7811eb6e9ef6b094/probreg-0.3.8.tar.gz",
            "yanked": false,
            "yanked_reason": null
        }
    ],
    "upload_time": "2024-05-11 06:42:17",
    "github": true,
    "gitlab": false,
    "bitbucket": false,
    "codeberg": false,
    "github_user": "neka-nat",
    "github_project": "probreg",
    "travis_ci": false,
    "coveralls": false,
    "github_actions": true,
    "lcname": "probreg"
}
neka-nat
Elapsed time: 0.24006s