Skip to content

Latest commit

 

History

History
108 lines (84 loc) · 2.5 KB

README.md

File metadata and controls

108 lines (84 loc) · 2.5 KB
Raw

Deep Object Co-Segmentation (DOCS) - PyTorch

This is the Pytorch implementation of our paper Deep Object Co-Segmentation published at ACCV18. For more information, you can visit the project page. You can also find our caffe version here.

DOCS Network

Requirements

  • Python 3
  • Pytorch >= 1.0
  • gcc (tested with gcc 8)

Tested on Ubuntu 20.04 with pytorch 1.5, CUDA 10.2 and gcc 8.0 .

NOTE: for using it with older version of pytorch (0.4.1) please check v1.0 .

Installation

git clone https://github.com/ohosseini/DOCS-pytorch.git
cd DOCS-pytorch
bash install.sh

Demo

First download the model from here and put it in DOCS-pytorch directory.

Then you can run the demo with

bash demo.sh

demo

For more information on how to apply the demo on other images you can check

python demo.py --help

Comparision with Caffe version

The numbers are compareable to the original ones in the paper which are generated using our main code (DOCS-caffe).

DOCS-caffe DOCS-pytorch
MSRC P 95.4 92.0
J 82.9 82.3
Internet P 93.5 92.9
J 72.6 72.0
iCoseg P 95.1 94.1
J 84.2 84.0

Citation

If you use this code, please cite our publication:

Deep Object Co-Segmentation, Weihao Li*, Omid Hosseini Jafari*, Carsten Rother, ACCV 2018.

@InProceedings{DOCS_ACCV18,
  title={Deep Object Co-Segmentation},
  author={Li, Weihao and Hosseini Jafari, Omid and Rother, Carsten},
  booktitle={ACCV},
  year={2018}
}

Acknowledgments

We used the Pytorch implementation of correlation layer from "Pytorch implementation of FlowNet 2.0" https://github.com/NVIDIA/flownet2-pytorch.