Decoupled Spatial-Temporal Transformer for Video Inpainting

By Rui Liu, Hanming Deng, Yangyi Huang, Xiaoyu Shi, Lewei Lu, Wenxiu Sun, Xiaogang Wang, Jifeng Dai, Hongsheng Li.

This repo is the official Pytorch implementation of Decoupled Spatial-Temporal Transformer for Video Inpainting.

Introduction

Usage

Prerequisites

Python >= 3.6
Pytorch >= 1.0 and corresponding torchvision (https://pytorch.org/)

Install

Clone this repo:

git clone https://github.com/ruiliu-ai/DSTT.git

Install other packages:

cd DSTT
pip install -r requirements.txt

Training

Dataset preparation

Download datasets (YouTube-VOS and DAVIS) into the data folder.

mkdir data

Training script

python train.py -c configs/youtube-vos.json

Test

Download pre-trained model into checkpoints folder.

mkdir checkpoints

Test script

python test.py -c checkpoints/dstt.pth -v data/DAVIS/JPEGImages/blackswan -m data/DAVIS/Annotations/blackswan

Citing DSTT

If you find DSTT useful in your research, please consider citing:

@article{Liu_2021_DSTT,
  title={Decoupled Spatial-Temporal Transformer for Video Inpainting},
  author={Liu, Rui and Deng, Hanming and Huang, Yangyi and Shi, Xiaoyu and Lu, Lewei and Sun, Wenxiu and Wang, Xiaogang and Li Hongsheng},
  journal={arXiv preprint arXiv:2104.06637},
  year={2021}
}

Acknowledement

This code relies heavily on the video inpainting framework from spatial-temporal transformer net.

hn18001/DSTT