Closed Kunyro closed 10 months ago
Hi @Kunyro,
My understanding is existing unsupervised, signal processing-based methods like POS and CHROM should work somewhat well, as well as maybe neural methods (such as DeepPhys) given you can utilize pseudo-labels, an existing dataset, or you can make your own curated dataset. I should note that you would have to make certain changes to this toolbox pipeline in order to achieve that, especially since this toolbox is focused on detecting and cropping to faces in particular for the rPPG task. I personally don't think this is too difficult, and if you have specific questions on how to re-purpose this toolbox's pipeline for other potential parts of the body for rPPG, we would be happy to help with any problems you might encounter as you dig into things.
I find multi-site physiological monitoring to be interesting and there is definitely a possibility we could add support for that kind of dataset later on in the future. I think you will find this recent paper to be an especially interesting read. The authors of that paper also released a dataset, the Multi-Site Physiological Monitoring (MSPM) dataset, which you should be able to get your hands on after filling out a form here.
All the best!
Alright, Very useful, thank you so much for the response. I will contact you if I have any more questions in the future
Thanks a lot
Hi,
I am interested in using neural network rPPG technology to detect peoples heart rate based on the palm of their hands. How would I go about doing that/is it even doable? Will something like DeepPhys work if trained on hands instead of faces? I am a bit of a noob but I am very interested in this technology and how to use it.
Thank you for the help