Home

Awesome

HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation

by Lukas Hoyer, Dengxin Dai, and Luc Van Gool

[ECCV22 Paper] [Extension Paper]

:bell: News:

Overview

Unsupervised domain adaptation (UDA) aims to adapt a model trained on synthetic data to real-world data without requiring expensive annotations of real-world images. As UDA methods for semantic segmentation are usually GPU memory intensive, most previous methods operate only on downscaled images. We question this design as low-resolution predictions often fail to preserve fine details. The alternative of training with random crops of high-resolution images alleviates this problem but falls short in capturing long-range, domain-robust context information.

Therefore, we propose HRDA, a multi-resolution training approach for UDA, that combines the strengths of small high-resolution crops to preserve fine segmentation details and large low-resolution crops to capture long-range context dependencies with a learned scale attention, while maintaining a manageable GPU memory footprint.

HRDA Overview

HRDA enables adapting small objects and preserving fine segmentation details. It significantly improves the state-of-the-art performance by 5.5 mIoU for GTA→Cityscapes and by 4.9 mIoU for Synthia→Cityscapes, resulting in an unprecedented performance of 73.8 and 65.8 mIoU, respectively.

UDA over time

The more detailed domain-adaptive semantic segmentation of HRDA, compared to the previous state-of-the-art UDA method DAFormer, can also be observed in example predictions from the Cityscapes validation set.

Demo

https://user-images.githubusercontent.com/1277888/181128057-27b8039f-a4c9-4f6d-9aa8-9b7f364d8921.mp4

Color Palette

HRDA can be further extended to domain generalization lifting the requirement of access to target images. Also in domain generalization, HRDA significantly improves the state-of-the-art performance by +4.2 mIoU.

For more information on HRDA, please check our [ECCV Paper] and the [Extension Paper].

If you find HRDA useful in your research, please consider citing:

@InProceedings{hoyer2022hrda,
  title={{HRDA}: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation},
  author={Hoyer, Lukas and Dai, Dengxin and Van Gool, Luc},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
  pages={372--391},
  year={2022}
}

@Article{hoyer2024domain,
  title={Domain Adaptive and Generalizable Network Architectures and Training Strategies for Semantic Image Segmentation},
  author={Hoyer, Lukas and Dai, Dengxin and Van Gool, Luc},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI)}, 
  year={2024},
  volume={46},
  number={1},
  pages={220-235},
  doi={10.1109/TPAMI.2023.3320613}
}

Comparison with SOTA UDA

HRDA significantly outperforms previous works on several UDA benchmarks. This includes synthetic-to-real adaptation on GTA→Cityscapes and Synthia→Cityscapes as well as clear-to-adverse-weather adaptation on Cityscapes→ACDC and Cityscapes→DarkZurich.

GTA→CS(val)Synthia→CS(val)CS→ACDC(test)CS→DarkZurich(test)
ADVENT [1]45.541.232.729.7
BDL [2]48.5--37.730.8
FDA [3]50.5--45.7--
DACS [4]52.148.3----
ProDA [5]57.555.5----
MGCDA [6]----48.742.5
DANNet [7]----50.045.2
DAFormer (Ours) [8]68.360.955.4*53.8*
HRDA (Ours)73.865.868.0*55.9*

* New results of our extension paper

References:

  1. Vu et al. "Advent: Adversarial entropy minimization for domain adaptation in semantic segmentation" in CVPR 2019.
  2. Li et al. "Bidirectional learning for domain adaptation of semantic segmentation" in CVPR 2019.
  3. Yang et al. "Fda: Fourier domain adaptation for semantic segmentation" in CVPR 2020.
  4. Tranheden et al. "Dacs: Domain adaptation via crossdomain mixed sampling" in WACV 2021.
  5. Zhang et al. "Prototypical pseudo label denoising and target structure learning for domain adaptive semantic segmentation" in CVPR 2021.
  6. Sakaridis et al. "Map-guided curriculum domain adaptation and uncertaintyaware evaluation for semantic nighttime image segmentation" in TPAMI, 2020.
  7. Wu et al. "DANNet: A one-stage domain adaptation network for unsupervised nighttime semantic segmentation" in CVPR, 2021.
  8. Hoyer et al. "DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation" in CVPR, 2022.

Comparison with SOTA Domain Generalization (DG)

HRDA and DAFormer significantly outperform previous works on domain generalization from GTA to real street scenes.

DG MethodCityscapesBDD100KMapillaryAvg.
IBN-Net [1,5]37.3734.2136.8136.13
DRPC [2]42.5338.7238.0539.77
ISW [3,5]37.2033.3635.5735.38
SAN-SAW [4]45.3341.1840.7742.43
SHADE [5]46.6643.6645.5045.27
DAFormer (Ours) [6]52.65*47.89*54.66*51.73*
HRDA (Ours)57.41*49.11*61.16*55.90*

* New results of our extension paper

References:

  1. Pan et al. "Two at once: Enhancing learning and generalization capacities via IBN-Net" in ECCV, 2018.
  2. Yue et al. "Domain randomization and pyramid consistency: Simulation-to-real generalization without accessing target domain data" ICCV, 2019.
  3. Choi et al. "RobustNet: Improving Domain Generalization in Urban-Scene Segmentation via Instance Selective Whitening" in CVPR, 2021.
  4. Peng et al. "Semantic-aware domain generalized segmentation" in CVPR, 2022.
  5. Zhao et al. "Style-Hallucinated Dual Consistency Learning for Domain Generalized Semantic Segmentation" in ECCV, 2022.
  6. Hoyer et al. "DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation" in CVPR, 2022.

Setup Environment

For this project, we used python 3.8.5. We recommend setting up a new virtual environment:

python -m venv ~/venv/hrda
source ~/venv/hrda/bin/activate

In that environment, the requirements can be installed with:

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

Please, download the MiT-B5 ImageNet weights provided by SegFormer from their OneDrive and put them in the folder pretrained/. Further, download the checkpoint of HRDA on GTA→Cityscapes and extract it to the folder work_dirs/.

Setup Datasets

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

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

Synthia (Optional): Please, download SYNTHIA-RAND-CITYSCAPES from here and extract it to data/synthia.

ACDC (Optional): Please, download rgb_anon_trainvaltest.zip and gt_trainval.zip from here and extract them to data/acdc. Further, please restructure the folders from condition/split/sequence/ to split/ using the following commands:

rsync -a data/acdc/rgb_anon/*/train/*/* data/acdc/rgb_anon/train/
rsync -a data/acdc/rgb_anon/*/val/*/* data/acdc/rgb_anon/val/
rsync -a data/acdc/gt/*/train/*/*_labelTrainIds.png data/acdc/gt/train/
rsync -a data/acdc/gt/*/val/*/*_labelTrainIds.png data/acdc/gt/val/

Dark Zurich (Optional): Please, download the Dark_Zurich_train_anon.zip and Dark_Zurich_val_anon.zip from here and extract it to data/dark_zurich.

BDD100K (Optional): Please, download the 10K Images and Segmentation from here and extract it to data/bdd100k.

Mapillary (Optional): Please, download the mapillary-vistas-dataset_public_v1.2.zip from here and extract it to data/mapillary.

The final folder structure should look like this:

HRDA
├── ...
├── data
│   ├── acdc (optional)
│   │   ├── gt
│   │   │   ├── train
│   │   │   ├── val
│   │   ├── rgb_anon
│   │   │   ├── train
│   │   │   ├── val
│   ├── bdd100k (optional)
│   │   ├── images/10k/val
│   │   ├── labels/sem_seg/masks/val
│   ├── cityscapes
│   │   ├── leftImg8bit
│   │   │   ├── train
│   │   │   ├── val
│   │   ├── gtFine
│   │   │   ├── train
│   │   │   ├── val
│   ├── dark_zurich (optional)
│   │   ├── gt
│   │   │   ├── val
│   │   ├── rgb_anon
│   │   │   ├── train
│   │   │   ├── val
│   ├── gta
│   │   ├── images
│   │   ├── labels
│   ├── mapillary (optional)
│   │   ├── validation/images
│   │   ├── validation/labels
│   ├── synthia (optional)
│   │   ├── 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
python tools/convert_datasets/mapillary.py data/mapillary/ --nproc 8

Testing & Predictions

The provided HRDA checkpoint trained on GTA→Cityscapes can be tested on the Cityscapes validation set using:

sh test.sh work_dirs/gtaHR2csHR_hrda_246ef

The predictions are saved for inspection to work_dirs/gtaHR2csHR_hrda_246ef/preds and the mIoU of the model is printed to the console. The provided checkpoint should achieve 73.79 mIoU. Refer to the end of work_dirs/gtaHR2csHR_hrda_246ef/20220215_002056.log for more information such as the class-wise IoU.

If you want to visualize the LR predictions, HR predictions, or scale attentions of HRDA on the validation set, please refer to test.sh for further instructions.

Training

For convenience, we provide an annotated config file of the final HRDA. A training job can be launched using:

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

The logs and checkpoints are stored in work_dirs/.

For the other experiments in our paper, we use a script to automatically generate and train the configs:

python run_experiments.py --exp <ID>

More information about the available experiments and their assigned IDs, can be found in experiments.py. The generated configs will be stored in configs/generated/.

When evaluating a model trained on Synthia→Cityscapes, please note that the evaluation script calculates the mIoU for all 19 Cityscapes classes. However, Synthia contains only labels for 16 of these classes. Therefore, it is a common practice in UDA to report the mIoU for Synthia→Cityscapes only on these 16 classes. As the Iou for the 3 missing classes is 0, you can do the conversion mIoU16 = mIoU19 * 19 / 16.

The results for Cityscapes→ACDC and Cityscapes→DarkZurich are reported on the test split of the target dataset. To generate the predictions for the test set, please run:

python -m tools.test path/to/config_file path/to/checkpoint_file --test-set --format-only --eval-option imgfile_prefix=labelTrainIds to_label_id=False

The predictions can be submitted to the public evaluation server of the respective dataset to obtain the test score.

Domain Generalization

HRDA/DAFormer for domain generalization (DG) is located on the DG branch, which can be checked out with:

git checkout dg

They can be trained for DG using:

python run_experiments.py --exp 50

For further details, please refer to experiment.py. The model is directly evaluated on Cityscapes during training with GTA data only. It can be additionally evaluated on BDD100K and Mapillary with tools/test.py:

python -m tools.test path/to/config_file path/to/checkpoint_file --eval mIoU --dataset BDD100K
python -m tools.test path/to/config_file path/to/checkpoint_file --eval mIoU --dataset Mapillary --eval-option efficient_test=True

Checkpoints

Below, we provide checkpoints of HRDA for different benchmarks. They come together with the log files of their training. As the results in the paper are provided as the mean over three random seeds, we provide the checkpoint with the median validation performance here.

The checkpoints come with the training logs. Please note that:

Framework Structure

This project is based on mmsegmentation version 0.16.0. For more information about the framework structure and the config system, please refer to the mmsegmentation documentation and the mmcv documentation.

The most relevant files for HRDA are:

Acknowledgements

HRDA is based on the following open-source projects. We thank their authors for making the source code publicly available.

License

This project is released under the Apache License 2.0, while some specific features in this repository are with other licenses. Please refer to LICENSES.md for the careful check, if you are using our code for commercial matters.