seenet's Introduction

SEE-Net

The Tensorflow implementation of "Order Matters: Shuffling Sequence Generation for Video Prediction" (BMVC2019) by Junyan Wang, Bingzhang Hu, Yang Long, Yu Guan.

Python packages

Python: 3.6.8
Tensorflow: 1.12.0
CUDA 9.0

Dataset & Pretrained Models

Make a directory ./data for saving models and a directory ./pretrained for pretrained models.

The pretrained SEENet Model can be downloaded from https://www.dropbox.com/sh/m4jrus3x7cjyh9t/AADxVp06scNlQDWLsVTHXHD1a?dl=0
The preprocessed KTH dataset can be downloaded from https://www.dropbox.com/sh/5wv2m9ov0usnikj/AAC6Fo3HMPIlXLho6pZKmNtUa?dl=0

Train

Make a directory ./models for saving models and a directory ./logs for saving logs.

To train motion and content features:

python3 ./src/main.py --train_feature True --test False

To train predict part:

python3 ./src/main.py --train_feature False --test False

The predicted samples can be seen in ./samples folder. The detailed arguments can be set up in ./src/args.py

Test

python3 ./src/main.py --test True

The test predicted samples can be seen in ./samples/test folder

Examples

Cite

If you use this code or reference our paper in your work please cite this publication as:

ArXiv Version:

@article{wang2019order,
  title={Order Matters: Shuffling Sequence Generation for Video Prediction},
  author={Wang, Junyan and Hu, Bingzhang and Long, Yang and Guan, Yu},
  journal={arXiv preprint arXiv:1907.08845},
  year={2019}
}

BMVC Version:

@inproceedings{wang2019order,
    author    = {Junyan Wang and
               Bingzhang Hu and
               Yang Long and
               Yu Guan},
    title     = {Order Matters: Shuffling Sequence Generation for Video Prediction},
    booktitle = {Proc. BMVA British Mach. Vis. Conf.},
    pages     = {275.1--275.14},
    year      = {2019}}