TreePartNet

KO | EN

gitlite — search

#javascript #ai #go #machine-learning #rust #deep-learning #linux #macos #awesome #awesome-list #vue #pytorch #security #openai

(★ 91)

This is the code repository implementing the paper "TreePartNet: Neural Decomposition of Point Clouds for 3D Tree Reconstruction".

파일 탐색기

최종 버전 다운로드 (.zip)

- epoch=167.ckpt
- epoch=28-val_loss=0.17-val_acc=0.951.ckpt
- PreProcess.py
- TreePartNet.py
- - __init__.py
  - _version.py
  - pointnet2_modules.py
  - pointnet2_utils.py
- MANIFEST.in
- setup.py
- formatTestData.py
- ply2h5_transformer.py
- post_process.py
- TreeDataset.py
README.md
requirements.txt
test.py
test_foliage.py
train.py
train_foliage.py
tree_data.png
TreePartNet.png

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# TreePartNet

This is the code repository implementing the paper "TreePartNet: Neural Decomposition of Point Clouds for 3D Tree Reconstruction".

![teaser](https://raw.githubusercontent.com/marktube/TreePartNet/main/TreePartNet.png)

## Dependency and Setup
The project is tested on Ubuntu 18.04 with cuda10.1.

Requirements:
- python==3.7
- pytorch==1.5.0
- pytorch-lightning==0.8.5

The PointNet++ pytorch implementation is modified from [Pointnet2_Pytorch](https://github.com/erikwijmans/Pointnet2_PyTorch). Install dependencies:
```
pip install -r requirements.txt
```

## Data
The gravity direction in tree point cloud is down along y-axis! All tree point cloud are normalized. see the code in utils for more details.

![data](https://raw.githubusercontent.com/marktube/TreePartNet/main/tree_data.png)

- Dataset for foliage segmentation: 16K points per tree, hdf5 format, [Download Link](https://mailsucaseducn-my.sharepoint.com/:f:/g/personal/liuyanchao18_mails_ucas_edu_cn/Eg3s4r5_7GZIneYgPeV7SNABLg6oVE7yJvgF38UJYKki6Q?e=zZiUlb), [Baiduyun for people in China(code:nva1)](https://pan.baidu.com/s/1XvFRE4OoHKFJVmThhlAiwg)
- Dataset for neural decomposition: 8K points per tree, hdf5 format, [Download Link](https://1drv.ms/u/s!Am5H_4z5WHyYa7TmZLatW6ZHMaE?e=W4toEN), [Baiduyun for people in China(code:4f5n)](https://pan.baidu.com/s/1Wu7fiziBKTqN6CyO-QFTFw)

## Training
After downloading the data and put them in data folder, the foliage segmentation network can be trained as
```
python train_foliage.py
```
and the TreePartNet can be trained using
```
python train.py
```
The hyperparameters can be modified in these 2 python files.

## Testing
The trained checkpoints can be found in dir *fckpt* and *ckpt*.
To predict foliage segmentation on test data set above:
```
python test_foliage.py
```
and neural decomposition:
```
python test.py
```

## Reference
If you find our work useful in your research, please cite us using the following BibTeX entry.
```
@article{TreePartNet21,
title={TreePartNet: Neural Decomposition of Point Clouds for 3D Tree Reconstruction},
author={Yanchao Liu and Jianwei Guo and Bedrich Benes and Oliver Deussen and Xiaopeng Zhang and Hui Huang},
journal={ACM Transactions on Graphics (Proceedings of SIGGRAPH ASIA)},
volume={40},
number={6},
pages={232:1--232:16},
year={2021},
} 
```

TreePartNet

파일 탐색기

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

TreePartNet

파일 탐색기

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README