Skip to content

KathyFeiyang/cs231n-project

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

44 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Temporally and Spatially Novel Video Frame Synthesis using 4D Video Autoencoder

Repository for the Stanford CS231N project: "Temporally and Spatially Novel Video Frame Synthesis using 4D Video Autoencoder." The repository includes implementation for the flow-enabled 4D Video Autoencoder, training scripts and test scripts. The code was adapted from zlai0/VideoAutoencoder of "Video Autoencoder: self-supervised disentanglement of 3D structure and motion" by Lai et al. (paper). Citation for this work can be found below:

@inproceedings{Lai21a,
        title={Video Autoencoder: self-supervised disentanglement of 3D structure and motion},
        author={Lai, Zihang and Liu, Sifei and Efros, Alexei A and Wang, Xiaolong},
        booktitle={ICCV},
        year={2021}
}

Prerequisites

Install dependences following the instructions here. These prerequisites are the same as the original Video Autoencoder.

Usage

flow_train.py is the entry point for training the 4D Video Autoencoder. A sample command is:

python flow_train.py --savepath log/model --resume mp3d.ckpt --dataset HMDB51 --epochs 100 --interval 1 --lr 2e-4 --bsize 1 --clip_length 4

flow_test_interpolate.py is the entry point for interpolating novel middle frames. A sample command is:

python flow_test_interpolate.py --savepath log/exp0_inter --resume flow_checkpoint.tar --dataset HMDB51 --interval 1 --video_limit 1000

After generating videos frames, such as through interpolation, you can use eval_syn_re10k.py to evaluate the quality of the generations using LIPIS, PSNR and SSIM. A sample command is:

python eval_syn_re10k.py --lpips log/interpolation_output/Videos/

Model checkpoint

We release the model checkpoint from our best performing model here.

License

This repository is made available to the public under the MIT license.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •