Awesome

Code for "Learning to Align Sequential Actions in the Wild" (CVPR 2022)

This repository is a TensorFlow implementation for the paper Learning to Align Sequential Actions in the Wild. Weizhe Liu, Bugra Tekin, Huseyin Coskun, Vibhav Vineet, Pascal Fua, Marc Pollefeys. CVPR. 2022 . If you use this code in your research, please cite the paper.

State-of-the-art methods for self-supervised sequential action alignment rely on deep networks that find correspondences across videos in time. They either learn frame-toframe mapping across sequences, which does not leverage temporal information, or assume monotonic alignment between each video pair, which ignores variations in the order of actions. As such, these methods are not able to deal with common real-world scenarios that involve background frames or videos that contain non-monotonic sequence of actions.

In this paper, we propose an approach to align sequential actions in the wild that involve diverse temporal variations. To this end, we propose an approach to enforce temporal priors on the optimal transport matrix, which leverages temporal consistency, while allowing for variations in the order of actions. Our model accounts for both monotonic and non-monotonic sequences and handles background frames that should not be aligned. We demonstrate that our approach consistently outperforms the stateof-the-art in self-supervised sequential action representation learning on four different benchmark datasets.

Figure 1: Alignment Priors: our approach, VAVA, enforces temporal priors to address non-monotonic frames and gracefully handles unmatched frames (e.g. crouching position in the first sequence), resulting in accurate alignment between sequences.

Installation

pip install -r requirements.txt

Dataset

 1. Download COIN Dataset from here

 2. We follow previous work to preprocess the dataset, please refer to this script for dataset preparation

VAVA Implementation

Please refer to "vava.py"

How to Use

The loss term we proposed is listed in "vava.py", precisely, the function "vava_loss" is the VAVA loss(L_VAVA in Eq.(13) of the paper) and the function "all_loss" is the final loss term(L_all in Eq.(17) of the paper), please carefully adjust the hyperparameters when applied to your own data.

Citing

@InProceedings{Liu_2022_CVPR,

author = {Weizhe Liu and Bugra Tekin and Huseyin Coskun and Vibhav Vineet and Pascal Fua and Marc Pollefeys},

title = {Learning to Align Sequential Actions in the Wild},

booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},

month = {June},

year = {2022}

}

Note

I already left EPFL and Microsoft, therefore I'm not able to reach my previous code/data. This is a quick implementation based on my memory and I do not even have the machine/data to test it, if you find any bug/typo with this code, please let me know.

Contact

For any questions regard this paper/code, please directly contact Weizhe Liu.