Awesome

PiPa: Pixel- and Patch-wise Self-supervised Learning for Domain Adaptative Semantic Segmentation

by Mu Chen, Zhedong Zheng, Yi Yang, and Tat-Seng Chua

Update at 22/07/2024, Our follow-up work DCF on UDA is accepted at ACM Multmedia 2024 as Oral Presentation (Score 6/5/5).

Update at 26/07/2023, The paper is accepted by ACM Multmedia 2023.

Update at 14/11/2022, ArXiv Version of PiPa is available.

Update at 13/12/2022, Code and Checkpoints release

Overview

Unsupervised Domain Adaptation (UDA) aims to enhance the generalization of the learned model to other domains. The domain-invariant knowledge is transferred from the model trained on labeled source domain, e.g., video game, to unlabeled target domains, e.g., real-world scenarios, saving annotation expenses. Existing UDA methods for semantic segmentation usually focus on minimizing the inter-domain discrepancy of various levels, e.g., pixels, features, and predictions, for extracting domain-invariant knowledge. However, the primary intra-domain knowledge, such as context correlation inside an image, remains underexplored. In an attempt to fill this gap, we propose a unified pixel- and patch-wise self-supervised learning framework, called PiPa, for domain adaptive semantic segmentation that facilitates intra-image pixel-wise correlations and patch-wise semantic consistency against different contexts. The proposed framework exploits the inherent structures of intra-domain images, which: (1) explicitly encourages learning the discriminative pixel-wise features with intraclass compactness and inter-class separability, and (2) motivates the robust feature learning of the identical patch against different contexts or fluctuations. <img src="resources/diagram.png">

Installation

In this project, we used python 3.8.5

conda create --name pipa -y python=3.8.5
conda activate pipa

Then, the requirements can be installed with:

pip install tensorboard

pip install -r requirements.txt -f https://download.pytorch.org/whl/torch_stable.html
pip install mmcv-full==1.3.7  # requires the other packages to be installed first

Further, download the MiT weights from SegFormer using the following script.

sh tools/download_checkpoints.sh

Prepare Data

Download [GTA5], [Synthia] and [Cityscapes] to run the basic code.

Cityscapes: download leftImg8bit_trainvaltest.zip and gt_trainvaltest.zip from here and extract them to data/cityscapes.

GTA: download all image and label packages from here and extract them to data/gta.

Synthia: download SYNTHIA-RAND-CITYSCAPES from here and extract it to data/synthia.

The data folder is structured as follows:

PiPa
├── ...
├── data
│   ├── cityscapes
│   │   ├── leftImg8bit
│   │   │   ├── train
│   │   │   ├── val
│   │   ├── gtFine
│   │   │   ├── train
│   │   │   ├── val
│   ├── gta
│   │   ├── images
│   │   ├── labels
│   ├── synthia
│   │   ├── RGB
│   │   ├── GT
│   │   │   ├── LABELS
├── ...

Data Preprocessing: Finally, please run the following scripts to convert the label IDs to the train IDs and to generate the class index for RCS:

python tools/convert_datasets/gta.py data/gta --nproc 8
python tools/convert_datasets/cityscapes.py data/cityscapes --nproc 8
python tools/convert_datasets/synthia.py data/synthia/ --nproc 8

Testing

We provide pretrained models below for PiPa based on hrda.

python -m tools.test path/to/config_file path/to/checkpoint_file --format-only --eval-option 

Training

python run_experiments.py --config configs/pipa/gtaHR2csHR_hrda.py

The logs and checkpoints are stored in work_dirs/.

Acknowledgements

We thank the authors of the following open-source projects for making the code publicly available.

• HRDA
• DAFormer
• MMSegmentation
• SegFormer
• DACS
• Seg-Uncertainty

Citation

If you find this work helpful to your research, please consider citing the paper:

@inproceedings{chen2023pipa,
  title={Pipa: Pixel-and patch-wise self-supervised learning for domain adaptative semantic segmentation},
  author={Chen, Mu and Zheng, Zhedong and Yang, Yi and Chua, Tat-Seng},
  booktitle={ACM Multimedia},
  year={2023}
}

@inproceedings{chen2024transferring,
  title={Transferring to Real-World Layouts: A Depth-aware Framework for Scene Adaptation},
  author={Chen, Mu and Zheng, Zhedong and Yang, Yi },
  booktitle={ACM Multimedia},
  year={2024}
}