# Learning Multiview 3D Point Cloud Registration
This repository provides code and data to train and evaluate the LMPCR, the first end-to-end algorithm for multiview registration of raw point clouds in a globally consistent manner. It represents the official implementation of the paper:
### [Learning Multiview 3D Point Cloud Registration (CVPR 2020).](https://arxiv.org/pdf/2001.05119.pdf)
\*[Zan Gojcic](https://www.ethz.ch/content/specialinterest/baug/institute-igp/geosensors-and-engineering-geodesy/en/people/scientific-assistance/zan-gojcic.html),\* [Caifa Zhou](https://ch.linkedin.com/in/caifa-zhou-7a461510b), [Jan D. Wegner](http://www.prs.igp.ethz.ch/content/specialinterest/baug/institute-igp/photogrammetry-and-remote-sensing/en/group/people/person-detail.html?persid=186562), [Leonidas J. Guibas](https://geometry.stanford.edu/member/guibas/), [Tolga Birdal](http://tbirdal.me/)\
|[EcoVision Lab ETH Zurich](https://prs.igp.ethz.ch/ecovision.html) | [Guibas Lab Stanford University](https://geometry.stanford.edu/index.html)|\
\* Equal contribution
We present a novel, end-to-end learnable, multiview 3D point cloud registration algorithm. Registration of multiple scans typically follows a two-stage pipeline: the initial pairwise alignment and the globally consistent refinement. The former is often ambiguous due to the low overlap of
neighboring point clouds, symmetries and repetitive scene parts. Therefore, the latter global refinement aims at establishing the cyclic consistency across multiple scans and helps in resolving the ambiguous cases. In this paper we propose, to the best of our knowledge, the first end-to-end algorithm for joint learning of both parts of this two-stage problem. Experimental evaluation on well accepted benchmark datasets shows that our approach outperforms the state-of-the-art by a significant margin, while being end-to-end trainable and computationally less costly. Moreover, we present detailed analysis and an ablation study that validate
the novel components of our approach.
![LM3DPCR](figures/LM3DPCR.jpg?raw=true)
### Citation
If you find this code useful for your work or use it in your project, please consider citing:
```shell
@inproceedings{gojcic2020LearningMultiview,
title={Learning Multiview 3D Point Cloud Registration},
author={Gojcic, Zan and Zhou, Caifa and Wegner, Jan D and Guibas, Leonidas J and Birdal, Tolga},
booktitle={International conference on computer vision and pattern recognition (CVPR)},
year={2020}
}
```
### Contact
If you have any questions or find any bugs, please let us know: Zan Gojcic {firstname.lastname@geod.baug.ethz.ch}
## Current state of the repository
Currently the repository contains only part of the code connected to the above mentioned publication and will be consistently updated in the course of the following days. The whole code will be available the following weeks.
**NOTE**: The published model is not the same as the model used in the CVPR paper. The results can therefore slightly differ from the ones in the paper. The models will be updated in the following days (with the fully converged ones).
## Instructions
The code was tested on Ubuntu 18.04 with Python 3.6, pytorch 1.5, CUDA 10.1.243, and GCC 7.
### Requirements
After cloning this repository, you can start by creating a virtual environment and installing the requirements by running:
```bash
conda create --name lmpr python=3.6
source activate lmpr
conda config --append channels conda-forge
conda install --file requirements.txt
conda install -c open3d-admin open3d=0.9.0.0
conda install -c intel scikit-learn
conda install pytorch torchvision cudatoolkit=10.1 -c pytorch
```
Our network uses [FCGF](https://github.com/chrischoy/FCGF) feature descriptor which is based on the [MinkowskiEnginge](https://github.com/StanfordVL/MinkowskiEngine) library for sparse tensors. In order to install Minkowski Engine run:
```bash
source activate lmpr
git clone https://github.com/StanfordVL/MinkowskiEngine.git
cd MinkowskiEngine
conda install numpy mkl-include
export CXX=g++-7; python setup.py install
cd ../
```
Finally, our network supports furthest point sampling, when sampling the interest points. To this end we require the [PointNet++](https://github.com/erikwijmans/Pointnet2_PyTorch/tree/master/pointnet2/models) library that can be installed as follows:
```bash
source activate lmpr
git clone https://github.com/erikwijmans/Pointnet2_PyTorch.git
cd Pointnet2_PyTorch
pip install -r requirements.txt
cd ../
```
## Pretrained models
We provide the pretrained models for [FCGF]((https://github.com/chrischoy/FCGF)) feature descriptor, our pairwise registration block, and jointly trained pairwise registration model. They can be downloaded using:
```bash
bash scripts/download_pretrained_models.sh
```
## Datasets
### Pairwise registration
In order to train the pairwise registration model, you have to download the full [3DMatch](http://3dmatch.cs.princeton.edu/) dataset. To (pre)-train the registration blocks you can either download the preprocessed dataset (~160GB) using
```bash
bash scripts/download_3DMatch_train.sh preprocessed
```
This dataset contains the pointwise correspondences established in the [FCGF](https://github.com/chrischoy/FCGF) feature space for all point cloud pairs of 3DMatch dataset.
We also provide the raw 3DMatch data (~4.5GB) and a script to generate the preprocess training data `./scripts/extract_data.py` that can be used either with the raw 3DMatch data or your personal dataset. The raw dataset can be downloaded using
```bash
bash scripts/download_3DMatch_train.sh raw
```
And then processed (extract FCGF feature descriptors, establish the correspondences, and save training data) using:
```bash
source activate lmpr
python ./scripts/extract_data.py \
--source_path ./data/train_data/ \
--target_path ./data/train_data/ \
--dataset 3d_match \
--model ./pretrained/fcgf/model_best.pth \
--voxel_size 0.025 \
--n_correspondences 20000 \
--inlier_threshold 0.05 \
--extract_features \
--extract_correspondences \
--extract_precomputed_training_data \
--with_cuda \
bash scripts/download_preprocessed_3DMatch.sh
```
## Training
Current release supports only training the pairwise registration network. The repository will be further updated in the next weeks. In order to train the pairwise registration network from scratch using the precomputed data run
```bash
source activate lmpr
python train.py ./configs/pairwise_registration/OANet.yaml
```
The training parameters can be set in `./configs/pairwise_registration/OANet.yaml`.
In order to fine tune the pairwise registration network in an end-to-end manner (including the FCGF block), the raw data has to be used. The code to sample the batches will be released in the next weeks.
## Evaluation
We provide the scripts for the automatic evaluation of our method on the 3DMatch and Redwood dataset. The results for our method can differ slightly from the results in the CVPR paper as we have retrained the model. Due to the different implementation the results for RANSAC might also differ slightly from
the results of the [official evaluation script](https://github.com/andyzeng/3dmatch-toolbox/tree/master/evaluation/geometric-registration).
### 3DMatch
To evaluate on 3DMatch you can either download the raw evaluation data (~450MB) using
```bash
bash scripts/download_3DMatch_eval.sh raw
```
or the processed data together with the results for our method and RANSAC (~2.7GB) using
```bash
bash scripts/download_3DMatch_eval.sh preprocessed
```
If you download the raw data you first have to process it (extract features and correspondences)
```bash
source activate lmpr
python ./scripts/extract_data.py \
--source_path ./data/eval_data/ \
--target_path ./data/eval_data/ \
--dataset 3d_match \
--model ./pretrained/fcgf/model_best.pth \
--voxel_size 0.025 \
--n_correspondences 5000 \
--extract_features \
--extract_correspo
没有合适的资源?快使用搜索试试~ 我知道了~
[CVPR2020]学习多视角3D点云配准_Python_Shell_下载.zip
![preview](https://csdnimg.cn/release/downloadcmsfe/public/img/white-bg.ca8570fa.png)
共46个文件
py:21个
txt:7个
yaml:5个
![preview-icon](https://csdnimg.cn/release/downloadcmsfe/public/img/scale.ab9e0183.png)
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 42 浏览量
2023-04-30
23:35:31
上传
评论
收藏 2.07MB ZIP 举报
温馨提示
[CVPR2020]学习多视角3D点云配准_Python_Shell_下载.zip
资源推荐
资源详情
资源评论
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![application/msword](https://img-home.csdnimg.cn/images/20210720083327.png)
![ckpt](https://img-home.csdnimg.cn/images/20210720083646.png)
收起资源包目录
![package](https://csdnimg.cn/release/downloadcmsfe/public/img/package.f3fc750b.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/TXT.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/TXT.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/TXT.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/TXT.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/TXT.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/TXT.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/TXT.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/JPG.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![folder](https://csdnimg.cn/release/downloadcmsfe/public/img/folder.005fa2e5.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
![file-type](https://csdnimg.cn/release/download/static_files/pc/images/minetype/UNKNOWN.png)
共 46 条
- 1
资源评论
![avatar-default](https://csdnimg.cn/release/downloadcmsfe/public/img/lazyLogo2.1882d7f4.png)
![avatar](https://profile-avatar.csdnimg.cn/default.jpg!1)
![avatar-vip](https://csdnimg.cn/release/downloadcmsfe/public/img/user-vip.1c89f3c5.png)
快撑死的鱼
- 粉丝: 1w+
- 资源: 9152
上传资源 快速赚钱
我的内容管理 展开
我的资源 快来上传第一个资源
我的收益
登录查看自己的收益我的积分 登录查看自己的积分
我的C币 登录后查看C币余额
我的收藏
我的下载
下载帮助
![voice](https://csdnimg.cn/release/downloadcmsfe/public/img/voice.245cc511.png)
![center-task](https://csdnimg.cn/release/downloadcmsfe/public/img/center-task.c2eda91a.png)
安全验证
文档复制为VIP权益,开通VIP直接复制
![dialog-icon](https://csdnimg.cn/release/downloadcmsfe/public/img/green-success.6a4acb44.png)