Video Object Segmentation using Space-Time Memory Networks

Seoung Wug Oh, Joon-Young Lee, Ning Xu, Seon Joo Kim

ICCV 2019

[paper]

- Requirements

• python 3.6
• pytorch 1.0.1.post2
• numpy, opencv, pillow

- How to Use

Download weights

Place it the same folder with demo scripts

wget -O STM_weights.pth "https://www.dropbox.com/s/mtfxdr93xc3q55i/STM_weights.pth?dl=1"

Run

DAVIS-2016 validation set (Single-object)

python eval_DAVIS.py -g '1' -s val -y 16 -D [path/to/DAVIS]

DAVIS-2017 validation set (Multi-object)

python eval_DAVIS.py -g '1' -s val -y 17 -D [path/to/DAVIS]

- Pre-computed Results

If you are not able to run our code but interested in our results. The pre-computed results will be helpful.

• [DAVIS-16-val]
• [DAVIS-17-val]
• [DAVIS-17-test-dev]
• [YouTube-VOS-18-valid]
• [YouTube-VOS-19-valid]

- Reference

If you find our paper and repo useful, please cite our paper. Thanks!

Video Object Segmentation using Space-Time Memory Networks
Seoung Wug Oh, Joon-Young Lee, Ning Xu, Seon Joo Kim
ICCV 2019

- Related Project

Fast Video Object Segmentation by Reference-Guided Mask Propagation
Seoung Wug Oh, Joon-Young Lee, Kalyan Sunkavalli, Seon Joo Kim
CVPR 2018

[paper] [github]

- Interactive VOS (Quantitative Evaluation)

A modified STM model is used for DAVIS Interactive VOS Challenge 2019 (https://davischallenge.org/challenge2019/interactive.html). If you are intersted in comparison with our interactive STM model, please use evaluation summary obtained from the DAVIS framework. [Download link (DAVIS-17-val)]. The timing is measured using a single 2080Ti GPU. We will make a demo software available soon. For the further questions, please contact me by E-mail.

- Terms of Use

This software is for non-commercial use only. The source code is released under the Attribution-NonCommercial-ShareAlike (CC BY-NC-SA) Licence (see this for details)