https://github.com/Evgeneus/Graph-Domain-Adaptaion

PyTorch code for the paper "Curriculum Graph Co-Teaching for Multi-target Domain Adaptation" (CVPR2021)
https://github.com/Evgeneus/Graph-Domain-Adaptaion

cdan computer-vision coteaching curriculum-learning domain-adaptation domainnet graph-convolutional-networks graph-domain-adaptation graph-neural-networks mtdap office pacs pytorch

Last synced: about 2 months ago
JSON representation

PyTorch code for the paper "Curriculum Graph Co-Teaching for Multi-target Domain Adaptation" (CVPR2021)

• Host: GitHub
• URL: https://github.com/Evgeneus/Graph-Domain-Adaptaion
• Owner: Evgeneus
• Created: 2021-03-24T19:15:44.000Z (over 3 years ago)
• Default Branch: master
• Last Pushed: 2021-07-06T10:19:54.000Z (about 3 years ago)
• Last Synced: 2024-04-17T01:18:13.615Z (5 months ago)
• Topics: cdan, computer-vision, coteaching, curriculum-learning, domain-adaptation, domainnet, graph-convolutional-networks, graph-domain-adaptation, graph-neural-networks, mtdap, office, pacs, pytorch
• Language: Python
• Homepage:
• Size: 3.08 MB
• Stars: 96
• Watchers: 3
• Forks: 18
• Open Issues: 2
• Metadata Files:
  • Readme: readme.md

Awesome Lists containing this project

README

        
## PyTorch code for the paper "Curriculum Graph Co-Teaching for Multi-target Domain Adaptation" (CVPR2021)

This repo presents PyTorch implementation of Multi-targe Graph Domain Adaptation framework from ["Curriculum Graph Co-Teaching for Multi-target Domain Adaptation" CVPR 2021](https://arxiv.org/abs/).

The framework is pivoted around two key concepts: *graph feature aggregation* and *curriculum learning* (see pipeline below or [project web-page](https://roysubhankar.github.io/graph-coteaching-adaptation/)).



## Results



## Environment

```

Python >= 3.6

PyTorch >= 1.8.1

```

To install dependencies run (line 1 for pip or line 2 for conda env): 

```

pip install -r requirements.txt

conda install --file requirements.txt

```

*Disclaimer.*  This code has been tested with cuda toolkit 10.2. Please install PyTorch as supported by your machine.

## Datasets

Four datasets are supported:

* Office-31 ([Kate Saenko et al., 2010](https://link.springer.com/content/pdf/10.1007/978-3-642-15561-1_16.pdf))

* Office-Home ([Venkateswara et al., 2017](https://arxiv.org/abs/1706.07522))

* PACS ([Li et al., 2017](https://arxiv.org/abs/1710.03077))

* DomainNet ([Peng et al., 2019](http://ai.bu.edu/M3SDA/))

To run this code, one must check if the txt file names in data/ are matching with the downloaded domain folders. For e.g.,  to run OfficeHome, the domain sub-folders should be art/, clipart/, product/ and real/ corresponding to art.txt, clipart.txt, product.txt and real.txt that can be found in the data/office-home/.

## Methods

* CDAN

* CDAN+E

## Commands

## Office-31

Run D-CGCT:

```

python src/main_dcgct.py \

        --method 'CDAN' \

        --encoder 'ResNet50' \

 	--dataset 'office31' \

 	--data_root [your office31 folder] \

 	--source 'webcam' \

 	--target 'dslr' 'amazon' \

 	--source_iters 200 \

 	--adapt_iters 3000 \

 	--finetune_iters 15000 \

 	--lambda_node 0.3 \

 	--output_dir 'office31-dcgct/webcam_rest/CDAN'

```

Run CGCT:

```

python src/main_cgct.py \

        --method 'CDAN' \

        --encoder 'ResNet50' \

 	--dataset 'office31' \

 	--data_root [your office31 folder] \

 	--source 'webcam' \

 	--target 'dslr' 'amazon' \

 	--source_iters 100 \

 	--adapt_iters 3000 \

 	--finetune_iters 15000 \

 	--lambda_node 0.1 \

 	--output_dir 'office31-cgct/webcam_rest/CDAN'

```

## Office-Home

```

python src/main_dcgct.py \

	--method 'CDAN' \

	--encoder 'ResNet50' \

	--dataset 'office-home' \

	--data_root [your OfficeHome folder] \

	--source 'art' \

	--target 'clipart' 'product' 'real' \

	--source_iters 500 \

	--adapt_iters 10000 \

	--finetune_iters 15000 \

	--lambda_node 0.3 \

	--output_dir 'officeHome-dcgct/art_rest/CDAN' 

```

```

python src/main_cgct.py \

	--method 'CDAN' \

	--encoder 'ResNet50' \

	--dataset 'office-home' \

	--data_root [your OfficeHome folder] \

	--source 'art' \

	--target 'clipart' 'product' 'real' \

	--source_iters 500 \

	--adapt_iters 5000 \

	--finetune_iters 15000 \

	--lambda_node 0.1 \

	--output_dir 'officeHome-cgct/art_rest/CDAN' 

```

## PACS

```

python src/main_dcgct.py \

	--method 'CDAN' \

	--encoder 'ResNet50' \

	--dataset 'pacs' \

	--data_root [your PACS folder] \

	--source 'photo' \

	--target 'cartoon' 'art_painting' 'sketch' \

	--source_iters 200 \

	--adapt_iters 3000 \

	--finetune_iters 15000  \

	--lambda_node 0.1 \

	--output_dir 'pacs-dcgct/photo_rest/CDAN'  

```

```

python src/main_cgct.py \

	--method 'CDAN' \

	--encoder 'ResNet50' \

	--dataset 'pacs' \

	--data_root [your PACS folder] \

	--source 'photo' \

	--target 'cartoon' 'art_painting' 'sketch' \

	--source_iters 200 \

	--adapt_iters 3000 \

	--finetune_iters 15000  \

	--lambda_node 0.1 \

	--output_dir 'pacs-cgct/photo_rest/CDAN'  

```

## DomainNet

```

python src/main_dcgct.py \

	--method 'CDAN' \

	--encoder 'ResNet101' \

	--dataset 'domain-net' \

	--data_root [your DomainNet folder] \

	--source 'sketch' \

	--target 'clipart' 'infograph' 'painting' 'real' 'quickdraw' \

	--source_iters 5000 \

	--adapt_iters 50000 \

	--finetune_iters 15000  \

	--lambda_node 0.3 \

	--output_dir 'domainNet-dcgct/sketch_rest/CDAN'

```

```

python src/main_cgct.py \

	--method 'CDAN' \

	--encoder 'ResNet101' \

	--dataset 'domain-net' \

	--data_root [your DomainNet folder] \

	--source 'sketch' \

	--target 'clipart' 'infograph' 'painting' 'real' 'quickdraw' \

	--source_iters 5000 \

	--adapt_iters 50000 \

	--finetune_iters 15000  \

	--lambda_node 0.3 \

	--output_dir 'domainNet-cgct/sketch_rest/CDAN'

```

## Citation

If you find our paper and code useful for your research, please consider citing our paper.

```

@inproceedings{roy2021curriculum,

  title={Curriculum Graph Co-Teaching for Multi-target Domain Adaptation},

  author={Roy, Subhankar and Krivosheev, Evgeny and Zhong, Zhun and Sebe, Nicu and Ricci, Elisa},

  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},

  year={2021}

}

```