Skip to content
/ CCM Public

[ECCV2020] Content-Consistent Matching for Domain Adaptive Semantic Segmentation

License

Notifications You must be signed in to change notification settings

Solacex/CCM

Repository files navigation

[ECCV20] Content-Consistent Matching for Domain Adaptive Semantic Segmentation

This is a PyTorch implementation of CCM.

News: GTA-4K list is available!

A smaller subset of GTA5 dataset that shares higher layout similarites with Cityscapes.

Prerequisites

To install requirements:

pip install -r requirements.txt
  • Python 3.6
  • GPU Memory: 24GB for the first stage(Source-only Model), and 12GB for the second stage
  • Pytorch 1.4.0

Getting Started

  1. Download the dataset GTA5 and Cityscapes.
  2. Download the ImageNet-pretrained Model [Link].
  3. Download the Source-only Model Link.

Training

To train the source-only model:

CUDA_VISIBLE_DEVICES=0 python so_run.py

To train the adaptation model:

CUDA_VISIBLE_DEVICES=0 python run.py

Evaluation

To perform evaluation on a multiple models under a directory:

python eval.py --frm your_dir 

To perform evaluation on single model:

python eval.py --frm model.pth --single

Citation

If you find it helpful, please consider citing:

@inproceedings{li2020content,
  title={Content-consistent matching for domain adaptive semantic segmentation},
  author={Li, Guangrui and Kang, Guoliang and Liu, Wu and Wei, Yunchao and Yang, Yi},
  booktitle={European Conference on Computer Vision},
  pages={440--456},
  year={2020},
  organization={Springer}
}

About

[ECCV2020] Content-Consistent Matching for Domain Adaptive Semantic Segmentation

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages