/GUI-IVOS

Pytorch GUI(demo) implementation of CVPR2021 paper and ECCV2020 paper, "Guided Interactive Video Object Segmentation Using Reliability-Based Attention Maps" and"Interactive Video Object Segmentation Using Global and Local Transfer Modules"

Primary LanguagePythonMIT LicenseMIT

Python 3.6

GUI for IVOS(interactive VOS) and GIS (Guided IVOS)

explain_qwerty GUI Implementation of

CVPR2021 paper "Guided Interactive Video Object Segmentation Using Reliability-Based Attention Maps"

ECCV2020 paper "Interactive Video Object Segmentation Using Global and Local Transfer Modules"

Githubs:
CVPR2021 / ECCV2020

Project Pages:
CVPR2021 / ECCV2020

Codes in this github:

  1. Real-world GUI evaluation on DAVIS2017 based on the DAVIS framework
  2. GUI for other videos

Prerequisite

  • cuda 11.0
  • python 3.6
  • pytorch 1.6.0
  • davisinteractive 1.0.4
  • numpy, cv2, PtQt5, and other general libraries of python3

Directory Structure

  • root/apps: QWidget apps.

  • root/checkpoints: save our checkpoints (pth extensions) here.

  • root/dataset_torch: pytorch datasets.

  • root/libs: library of utility files.

  • root/model_CVPR2021 : networks and GUI models for CVPR2021

  • root/model_ECCV2020 : networks and GUI models for ECCV2020

  • root/eval_GIS_RS1.py : DAVIS2017 evaluation based on the DAVIS framework.

  • root/eval_GIS_RS4.py : DAVIS2017 evaluation based on the DAVIS framework.

  • root/eval_IVOS.py : DAVIS2017 evaluation based on the DAVIS framework.

  • root/IVOS_demo_customvideo.py : GUI for custom videos

Instruction

To run

  1. Edit eval_GIS_RS1.py``eval_GIS_RS4.py``eval_IVOS.py``IVOS_demo_customvideo.py to set the directory of your DAVIS2017 dataset and other configurations.
  2. Download our parameters and place the file as root/checkpoints/GIS-ckpt_standard.pth.
  3. Run eval_GIS_RS1.py``eval_GIS_RS4.py``eval_IVOS.py for real-world GUI evaluation on DAVIS2017 or
  4. Run IVOS_demo_customvideo.py to apply our method on the other videos

To use

explain_qwerty

Left click for the target object and right click for the background.

  1. Select any frame to interact by dragging the slidder under the main image
  2. Give interaction
  3. Run VOS
  4. Find worst frame and reinteract. - For GIS, a candidate frame(RS1) or candidate frames(RS4) are given
  5. Iterate until you get satisfied with VOS results.
  6. By selecting satisfied button, your evaluation result (consumed time and frames) will be recorded on root/results.

Reference

Please cite our paper if the implementations are useful in your work:

@Inproceedings{
Yuk2021GIS,
title={Guided Interactive Video Object Segmentation Using Reliability-Based Attention Maps},
author={Yuk Heo and Yeong Jun Koh and Chang-Su Kim},
booktitle={CVPR},
year={2021},
url={https://openaccess.thecvf.com/content/CVPR2021/papers/Heo_Guided_Interactive_Video_Object_Segmentation_Using_Reliability-Based_Attention_Maps_CVPR_2021_paper.pdf}
}

@Inproceedings{
Yuk2020IVOS,
title={Interactive Video Object Segmentation Using Global and Local Transfer Modules},
author={Yuk Heo and Yeong Jun Koh and Chang-Su Kim},
booktitle={ECCV},
year={2020},
url={https://openreview.net/forum?id=bo_lWt_aA}
}

Our real-world evaluation demo is based on the GUI of IPNet:

@Inproceedings{
Oh2019IVOS,
title={Fast User-Guided Video Object Segmentation by Interaction-and-Propagation Networks},
author={Seoung Wug Oh and Joon-Young Lee and Seon Joo Kim},
booktitle={CVPR},
year={2019},
url={https://openaccess.thecvf.com/content_ICCV_2019/papers/Oh_Video_Object_Segmentation_Using_Space-Time_Memory_Networks_ICCV_2019_paper.pdf}
}