Made to Order: Discovering monotonic temporal changes via self-supervised video ordering

Charig Yang, Weidi Xie, Andrew Zisserman

ECCV, 2024 (Oral Presentation)

Visual Geometry Group, Department of Engineering Science, University of Oxford

Requirements

pytorch, opencv, einops, tensorboardX

How to use

To get started,

python main.py

This should train the model on MNIST under default settings. You may visualise the training and attribution maps on Tensorboard.

We have included a instructions on how to train on several datsets (RDS, MNIST and SVHN). Check main.py. The dataset should be downloaded automatically on the first run (or created on the fly, as in RDS).

Other datasets, see https://drive.google.com/file/d/1y0_2H_oCT4ixIGhmK64AlJzIYxHxId4W/view?usp=sharing

To run this on your own dataset, simply create a dataloader of the same nature.

Citation

If you find this repository helpful, please consider citing our work:

@InProceedings{yang2024made,
      title={Made to Order: Discovering monotonic temporal changes via self-supervised video ordering}, 
      author={Charig Yang and Weidi Xie and Andrew Zisserman},
      booktitle={ECCV},
      year={2024},
}