lightas / CVPR2020_MANet

This is the pytorch implementation of the CVPR2020 paper "Memory aggregation networks for efficient interactive video object segmentation".

CVPR2020 Memory aggregation networks for efficient interactive video object segmentation

This is the pytorch implementation of the CVPR2020 paper "Memory aggregation networks for efficient interactive video object segmentation".

Preparation

Dependencies

• Python 3.7
• Pytorch 1.0
• Numpy
• tensorboardX
• davisinteractive (Please refer to this link)

Pretrained model

Download deeplabV3+ model pretrained on COCO to this repo.

Dataset

Download DAVIS2017 and scribbles into one folder. Please refer to DAVIS.

If you need the file "DAVIS2017/ImageSets/2017/v_a_l_instances.txt", please refer to the link https://drive.google.com/file/d/1aLPaQ_5lyAi3Lk3d2fOc_xewSrfcrQlc/view?usp=sharing

Train and Test

sh run_local.sh

Evaluation

You can download our model and decompress it for evaluation.

Citation

Please cite this paper in your publications if it helps your research:

@inproceedings{miao2020memory,
  title={Memory aggregation networks for efficient interactive video object segmentation},
  author={Miao, Jiaxu and Wei, Yunchao and Yang, Yi},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={10366--10375},
  year={2020}
}