https://github.com/wxinlong/asis

Associatively Segmenting Instances and Semantics in Point Clouds, CVPR 2019
https://github.com/wxinlong/asis

deep-learning instance-segmentation point-cloud semantic-segmentation

Last synced: 3 days ago
JSON representation

Associatively Segmenting Instances and Semantics in Point Clouds, CVPR 2019

• Host: GitHub
• URL: https://github.com/wxinlong/asis
• Owner: WXinlong
• License: mit
• Created: 2019-02-26T02:46:20.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-04-27T12:22:24.000Z (almost 6 years ago)
• Last Synced: 2025-01-16T18:21:02.223Z (10 days ago)
• Topics: deep-learning, instance-segmentation, point-cloud, semantic-segmentation
• Language: Python
• Homepage:
• Size: 1.64 MB
• Stars: 256
• Watchers: 3
• Forks: 65
• Open Issues: 17
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Associatively Segmenting Instances and Semantics in Point Clouds

The full paper is available at: https://arxiv.org/abs/1902.09852.

Qualitative results of ASIS on the S3DIS and vKITTI test fold:

![](misc/s3dis_asis.png)

![](misc/vkitti_asis.png)

## Overview

![](misc/fig.png)

## Dependencies

The code has been tested with Python 2.7 on Ubuntu 14.04.

*  [TensorFlow](https://www.tensorflow.org/)

*  h5py

## Data and Model

* Download 3D indoor parsing dataset ([S3DIS Dataset](https://docs.google.com/forms/d/e/1FAIpQLScDimvNMCGhy_rmBA2gHfDu3naktRm6A8BPwAWWDv-Uhm6Shw/viewform?c=0&w=1)). Version 1.2 of the dataset is used in this work.

``` bash

python collect_indoor3d_data.py

python gen_h5.py

cd data && python generate_input_list.py

cd ..

```

* (optional) Trained model can be downloaded from [here](https://drive.google.com/open?id=1UF2nfXdWTOa1iXXmD54_c09rM7pr-kMK).

## Usage

* Compile TF Operators

  Refer to [PointNet++](https://github.com/charlesq34/pointnet2)

* Training

``` bash

cd models/ASIS/

ln -s ../../data .

sh +x train.sh 5

```

* Evaluation

``` bash

python eval_iou_accuracy.py

```

Note: We test on Area5 and train on the rest folds in default. 6 fold CV can be conducted in a similar way.

## Citation

If our work is useful for your research, please consider citing:

	@inproceedings{wang2019asis,

		title={Associatively Segmenting Instances and Semantics in Point Clouds},

		author={Wang, Xinlong and Liu, Shu and Shen, Xiaoyong and Shen, Chunhua, and Jia, Jiaya},

		booktitle={CVPR},

		year={2019}

	}

## Acknowledgemets

This code largely benefits from following repositories:

[PointNet++](https://github.com/charlesq34/pointnet2),

[SGPN](https://github.com/laughtervv/SGPN),

[DGCNN](https://github.com/WangYueFt/dgcnn) and

[DiscLoss-tf](https://github.com/hq-jiang/instance-segmentation-with-discriminative-loss-tensorflow)