carrierlxk / COSNet

See More, Know More: Unsupervised Video Object Segmentation with Co-Attention Siamese Networks (CVPR19)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

COSNet

Code for CVPR 2019 paper:

See More, Know More: Unsupervised Video Object Segmentation with Co-Attention Siamese Networks

Xiankai Lu, Wenguan Wang, Chao Ma, Jianbing Shen, Ling Shao, Fatih Porikli


🆕

Our group co-attention achieves a further performance gain (81.1 mean J on DAVIS-16 dataset), related codes have also been released.

The pre-trained model, testing and training code:

Quick Start

Testing

  1. Install pytorch (version:1.0.1).

  2. Download the pretrained model. Run 'test_coattention_conf.py' and change the davis dataset path, pretrainde model path and result path.

  3. Run command: python test_coattention_conf.py --dataset davis --gpus 0

  4. Post CRF processing code comes from: https://github.com/lucasb-eyer/pydensecrf.

The pretrained weight can be download from GoogleDrive or BaiduPan, pass code: xwup.

The segmentation results on DAVIS, FBMS and Youtube-objects can be download from DAVIS_benchmark(https://davischallenge.org/davis2016/soa_compare.html) or GoogleDrive or BaiduPan, pass code: q37f.

The youtube-objects dataset can be downloaded from here and annotation can be found here.

The FBMS dataset can be downloaded from here.

Training

  1. Download all the training datasets, including MARA10K and DUT saliency datasets. Create a folder called images and put these two datasets into the folder.

  2. Download the deeplabv3 model from GoogleDrive. Put it into the folder pretrained/deep_labv3.

  3. Change the video path, image path and deeplabv3 path in train_iteration_conf.py. Create two txt files which store the saliency dataset name and DAVIS16 training sequences name. Change the txt path in PairwiseImg_video.py.

  4. Run command: python train_iteration_conf.py --dataset davis --gpus 0,1

Citation

If you find the code and dataset useful in your research, please consider citing:

@InProceedings{Lu_2019_CVPR,  
author = {Lu, Xiankai and Wang, Wenguan and Ma, Chao and Shen, Jianbing and Shao, Ling and Porikli, Fatih},  
title = {See More, Know More: Unsupervised Video Object Segmentation With Co-Attention Siamese Networks},  
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},  
year = {2019}  
}
@article{lu2020_pami,
  title={Zero-Shot Video Object Segmentation with Co-Attention Siamese Networks},
  author={Lu, Xiankai and Wang, Wenguan and Shen, Jianbing and Crandall, David and Luo, Jiebo},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2020},
  publisher={IEEE}
}

Other related projects/papers:

Saliency-Aware Geodesic Video Object Segmentation (CVPR15)

Learning Unsupervised Video Primary Object Segmentation through Visual Attention (CVPR19)

Any comments, please email: carrierlxk@gmail.com

About

See More, Know More: Unsupervised Video Object Segmentation with Co-Attention Siamese Networks (CVPR19)


Languages

Language:Python 100.0%