Summary of Non-contrastive Unsupervised Learning Of Physiological Signals From Video, by Jeremy Speth et al.
Non-Contrastive Unsupervised Learning of Physiological Signals from Video
by Jeremy Speth, Nathan Vance, Patrick Flynn, Adam Czajka
First submitted to arxiv on: 14 Mar 2023
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This research paper introduces an unsupervised learning framework that can extract subtle periodic signals from RGB videos, revolutionizing remote health monitoring at low cost. The current state-of-the-art solutions for remote photoplethysmography (rPPG) rely on deep learning models trained and evaluated on benchmark datasets with ground truth data from contact-PPG sensors. In contrast, this framework uses a non-contrastive approach to learn signal regression without relying on labelled video data. By minimizing the assumptions of periodicity and finite bandwidth, the model can discover the blood volume pulse directly from unlabelled videos. The authors demonstrate the effectiveness of their approach by training robust pulse rate estimators using unlabelled video data not specifically designed for rPPG. With its impressive empirical results and limited inductive biases, this framework has the potential to discover other periodic signals from video, enabling multiple physiological measurements without the need for ground truth signals. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine being able to monitor your health remotely using just a smartphone or camera. This research makes that possible by developing a new way to extract subtle signals from videos. These signals are like a heartbeat or breathing pattern, and they can be used to track important health metrics. The current best approaches for doing this rely on special sensors that require contact with the body. But this new method uses just regular video cameras and doesn’t need any special equipment. The researchers tested their approach using unlabelled videos and were able to accurately detect heartbeats and other signals. This could lead to a lot of new possibilities for remote health monitoring, making it easier and more convenient for people to track their health. |
Keywords
* Artificial intelligence * Deep learning * Regression * Unsupervised
