weixiong-ur / mdgan

official code of CVPR'18 paper "learning to generate time-lapse videos using multi-stage dynamic generative adversarial networks"

Learning to Generate Time-lapse Videos Using Multi-stage Dynamic Generative Adversarial Networks

This is the official code of the CVPR 2018 PAPER.

CVPR 2018 PAPER | Project Page | Dataset

Usage

1. Requirements:
   • download our time-lapse dataset
   • python2.7
   • pytorch 0.3.0 or 0.3.1
   • ffmpeg
2. Testing:
   • download our pretrained models
   • run python test.py --cuda --testf your_test_dataset_folder
3. Sample outputs:
   • in ./sample_outputsthere are mp4 files which are generated on my machine.

Citing

@InProceedings{Xiong_2018_CVPR,
author = {Xiong, Wei and Luo, Wenhan and Ma, Lin and Liu, Wei and Luo, Jiebo},
title = {Learning to Generate Time-Lapse Videos Using Multi-Stage Dynamic
Generative Adversarial Networks},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition
(CVPR)},
month = {June},
year = {2018}
}