Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/speedinghzl/dsrg
Weakly-Supervised Semantic Segmentation Network with Deep Seeded Region Growing (CVPR 2018).
https://github.com/speedinghzl/dsrg
caffe pascal-voc segmentation-network semantic-segmentation
Last synced: 15 days ago
JSON representation
Weakly-Supervised Semantic Segmentation Network with Deep Seeded Region Growing (CVPR 2018).
- Host: GitHub
- URL: https://github.com/speedinghzl/dsrg
- Owner: speedinghzl
- License: mit
- Created: 2018-03-15T12:06:52.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2018-12-30T10:50:38.000Z (almost 6 years ago)
- Last Synced: 2023-11-07T15:23:19.923Z (about 1 year ago)
- Topics: caffe, pascal-voc, segmentation-network, semantic-segmentation
- Language: Python
- Homepage:
- Size: 228 KB
- Stars: 247
- Watchers: 12
- Forks: 36
- Open Issues: 15
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Weakly-Supervised Semantic Segmentation Network with Deep Seeded Region Growing (CVPR2018)
By [Zilong Huang](http://speedinghzl.github.io), [Xinggang Wang](http://mclab.eic.hust.edu.cn/~xwang/index.htm), [Jiasi Wang](https://github.com/JiasiWang), [Wenyu Liu](http://mclab.eic.hust.edu.cn/MCWebDisplay/PersonDetails.aspx?Name=Wenyu%20Liu) and [Jingdong Wang](https://jingdongwang2017.github.io/).
This code is a implementation of the weakly-supervised semantic segmentation experiments in the paper [DSRG](http://openaccess.thecvf.com/content_cvpr_2018/papers/Huang_Weakly-Supervised_Semantic_Segmentation_CVPR_2018_paper.pdf). The code is developed based on the Caffe framework.
## Introduction
Overview of the proposed approach. The Deep Seeded Region Growing module takes the seed cues and segmentation map as input to produces latent pixel-wise supervision which is more accurate and more complete than seed cues. Our method iterates between refining pixel-wise supervision and optimizing the parameters of a segmentation network.
### License
DSRG is released under the MIT License (refer to the LICENSE file for details).
### Citing DSRG
If you find DSRG useful in your research, please consider citing:
@inproceedings{huang2018dsrg,
title={Weakly-Supervised Semantic Segmentation Network with Deep Seeded Region Growing},
author={Huang, Zilong and Wang, Xinggang and Wang, Jiasi and Liu, Wenyu and Wang, Jingdong},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={7014--7023},
year={2018}
}
## Installing dependencies
* Python packages:
```bash
$ pip install -r python-dependencies.txt
```
* **caffe (deeplabv2 version)**: deeplabv2 caffe installation instructions are available at `https://bitbucket.org/aquariusjay/deeplab-public-ver2`. Note, you need to compile **caffe** with python wrapper and support for python layers. Then add the caffe python path into [training/tools/findcaffe.py](https://github.com/speedinghzl/DSRG/blob/master/training/tools/findcaffe.py#L21).
* Fully connected CRF wrapper (requires the **Eigen3** package).
```bash
$ pip install CRF/
```
## Training the DSRG model
* Go into the training directory:
```bash
$ cd training
$ mkdir localization_cues
```
* Download the initial [VGG16](https://drive.google.com/open?id=1nq49w4os6BZ1JcrM4xqZKZh1wR3-32wi) model pretrained on Imagenet and put it in *training/* folder.
* Download CAM [seed](https://drive.google.com/open?id=1cHyhjul9srPlwcl4xqrYR9MwzhFGwKXU) and put it in *training/localization_cues* folder. We use [CAM](http://cnnlocalization.csail.mit.edu/) for localizing the foreground seed classes and utilize the saliency detection technology [DRFI](http://supermoe.cs.umass.edu/~hzjiang/drfi/) for localizing background seed. We provide the python interface to DRFI [here](https://github.com/speedinghzl/drfi_cpp) for convenience if you want to generate the seed by yourself.
```bash
$ cd training/experiment/seed_mc
$ mkdir models
```
* Set *root_folder* parameter in **train-s.prototxt, train-f.prototxt** and *PASCAL_DIR* in **run-s.sh** to the directory with **PASCAL VOC 2012** images
* Run:
```bash
$ bash run.sh
```
The trained model will be created in `models`
## Acknowledgment
This code is heavily borrowed from [SEC](https://github.com/kolesman/SEC).