# ContextPrior

**Repository Path**: AI52CV/ContextPrior

## Basic Information

- **Project Name**: ContextPrior
- **Description**: 场景分割
Implementation for "Context Prior for Scene Segmentation"
代码原地址：https://github.com/ycszen/ContextPrior
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 2
- **Forks**: 0
- **Created**: 2021-04-06
- **Last Updated**: 2022-01-17

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Context Prior for Scene Segmentation (CVPR2020)
Implementation for **[CPNet](https://arxiv.org/abs/2004.01547)**.

[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/context-prior-for-scene-segmentation/semantic-segmentation-on-ade20k)](https://paperswithcode.com/sota/semantic-segmentation-on-ade20k?p=context-prior-for-scene-segmentation)

[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/context-prior-for-scene-segmentation/scene-understanding-on-ade20k-val)](https://paperswithcode.com/sota/scene-understanding-on-ade20k-val?p=context-prior-for-scene-segmentation)

[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/context-prior-for-scene-segmentation/semantic-segmentation-on-pascal-context)](https://paperswithcode.com/sota/semantic-segmentation-on-pascal-context?p=context-prior-for-scene-segmentation)

[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/context-prior-for-scene-segmentation/semantic-segmentation-on-cityscapes)](https://paperswithcode.com/sota/semantic-segmentation-on-cityscapes?p=context-prior-for-scene-segmentation)

![Network](figs/network.png)

![Affinity](figs/affinity-map.png) | ![Aggregation](figs/aggregation-module.png)
:-------------------------:|:-------------------------:
## News
- A simple version of the core codes has been released. It is based on mmsegmentation. You can integrate the codes into mmseg to run the experiments. Recently I am busy with job hunting and PhD thesis. After this peroid of time, I will release the complete version.
- We achieve **46.3%** mean IoU on ADE20K *val* set, **53.9%** mIoU on PASCAL-Context, and **81.3%** mIoU on Cityscapes.

## Performance
![ADE](figs/ade.png)|![Pascal-Context](figs/pascal-cxt.png)|![Cityscapes](figs/cityscapes.png)
:-------------------------:|:-------------------------:|:-------------------------:

## Citation
Please consider citing the ContextPrior in your publications if it helps your research. 
```
@inproceedings{Yu-CVPR-ContextPrior-2020,
  title={Context Prior for Scene Segmentation},
  author={Yu, Changqian and Wang, Jingbo and Gao, Changxin and Yu, Gang and Shen, Chunhua and Sang, Nong},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2020},
}
```