https://github.com/Jin-Ying/GTA-Seg

Code for GTA-Seg (NeurIPS2022)
https://github.com/Jin-Ying/GTA-Seg

Last synced: about 1 month ago
JSON representation

Code for GTA-Seg (NeurIPS2022)

• Host: GitHub
• URL: https://github.com/Jin-Ying/GTA-Seg
• Owner: Jin-Ying
• License: apache-2.0
• Created: 2022-10-10T08:59:13.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2023-01-02T07:06:45.000Z (almost 2 years ago)
• Last Synced: 2024-08-03T01:11:51.643Z (5 months ago)
• Language: Python
• Size: 303 KB
• Stars: 37
• Watchers: 1
• Forks: 6
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• Awesome-Semi-Supervised-Semantic-Segmentation - Code - Ying/GTA-Seg)|[Paper](https://arxiv.org/pdf/2301.07340)| (2022)

README

        
# GTA-Seg

Code release for [Semi-Supervised Semantic Segmentation via Gentle Teaching Assistant], NeurIPS 2022.

## Installation

```bash

cd GTA-Seg

conda create -n gta python=3.6.9

conda activate gta

pip install -r requirements.txt

pip install pip install torch==1.8.1+cu102 torchvision==0.9.1+cu102 -f https://download.pytorch.org/whl/torch_stable.html

```

### Prepare Data

  For PASCAL VOC 2012

Download "VOCtrainval_11-May-2012.tar" from: http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar.

And unzip the files to folder ```data``` and make the dictionary structures as follows:

```angular2html

data/VOC2012

├── Annotations

├── ImageSets

├── JPEGImages

├── SegmentationClass

├── SegmentationClassAug

└── SegmentationObject

```

### Prepare Pretrained Backbone

Before training, please download ResNet101 pretrained on ImageNet-1K from one of the following:

  - [Google Drive](https://drive.google.com/file/d/1nzSX8bX3zoRREn6WnoEeAPbKYPPOa-3Y/view?usp=sharing)

  - [Baidu Drive](https://pan.baidu.com/s/1FDQGlhjzQENfPp4HTYfbeA) Fetch Code: 3p9h

After that, modify ```model_urls``` in ```models/resnet.py``` to ``````

### Train a Semi-Supervised Model

We can train a model on PASCAL VOC 2012 with ```183``` labeled data for supervision by:

```bash

cd experiments/pascal/183/ours

# use slurm

sh slurm_train.sh   

# or use torch.distributed.launch

# sh train.sh  

```

## Acknowledgement

We reproduce our work based on **U2PL**.

- U2PL: https://github.com/Haochen-Wang409/U2PL

Sincere gratitude to their work.

## Citation

```bibtex

@inproceedings{jin2022semi,

    title={Semi-Supervised Semantic Segmentation via Gentle Teaching Assistant},

    author={Jin, Ying and Wang, Jiaqi and Lin, Dahua},

    booktitle={Advances in Neural Information Processing Systems},

    year={2022}

}

```

## Contact

- Ying Jin, [email protected]