Covid-19 and Cough. This AI Predicts if You Have the Infection || Paper Explained.

[@MatteoSammarco]

Awsome presentation Valerio! Just one question: in the barplot for results, the baseline should be the one without pre-training weights, but it appears in the "No pre-trainint" and "pre-training" with a different value of discrimination.

[@andrehoffman6887]

Valerio, your videos/analysis are truly a great contribution. Thank you. Do you believe that using CNN's + the correct Biomarkers Models we could reach significant results in the whole area of Sound associated with the human body? And here I'd like to specifically mention the Snoring Sound. Does it make sense to reach out a Lab or 'Sleeping Center' and maybe try something related?

[@ValerioVelardoTheSoundofAI]

I'm not an expert in the area of snoring sound detection, so please take my comment with a pinch of salt. In general, I think injecting a degree of domain knowledge in black-box algorithms can be useful to improve performance. It may be interesting to reach out to a sleeping centre and collaborate with domain experts to create something similar to the models proposed in the paper I discussed.

[@markusbuchholz3518]

The research is very promising and probably worth attention beside your contribution. Generally, science and research should mainly contribute and focus to improve standard of living of all inhabitants of our planet (people and great animals) . It happens of course but still unfortunately the research is not boosted enough (financed) and promoted as it should. I hope open sources research projects will definitely change this trend. Anyway I hope also the COVID-19 will be killed to death! Thank you all and keep safe.

[@ValerioVelardoTheSoundofAI]

That's a great point. I also think that is up to the scientists to help spread awareness about the amazing work they often do in order to mobilise the population and pressure governments to fund research.

[@markusbuchholz3518]

@@ValerioVelardoTheSoundofAI Thanks for feedback. I hope it will change since new US president promised ... and all will follow ...

[@tirellarzu]

Is the open voice brain model open source?

[@ValerioVelardoTheSoundofAI]

I don't know.

[@jayeshdeshmukh9503]

do you have code of this?

[@agastoump]

How does the AI exclude that this is not one of the other respiratory diseases and not covid. So if you have 20000 people who all have different diseases with lung problems and similar symptoms how can this accurately find the specific disease? So if you have 20000 people with similar symptoms cough etc etc and only 1 person with covid will this find that 1 person that has covid or will it flag everyone as a covid patient?

[@myspam4194]

about the dataset, is it public ?

[@ValerioVelardoTheSoundofAI]

I don't know this, but I don't recall finding a link to the dataset in the paper.

[@muhammadazam4062]

Dataset is public , I am also working on it if anyone wants to work with me?

[@ariannaarienzo8479]

@@muhammadazam4062 hi Azam, i am also interested, could you share the database? [e-mail]

[@gabrielparra5562]

A very detailed explanation. Thank s Valerio. Your work is perfect to stimulate the research in other areas as in music as well.

[@ValerioVelardoTheSoundofAI]

Thank you Gabriel!

[@muhammadazam4062]

I have the dataset available , anyone wants to work with me?

[@sfgs2693]

Hello. I've downloaded coughvid dataset which includes 15000+ audio i need. I successfully transform these stuff into mel spectrogram and using imagefolder to split to different category. Begin to training,but i have failed dozen of times. Now what should I do?

[@TheMagicmagic290]

Not trying to be negative since the methodology is very interesting, but their data is pretty bad (46% of the positives and 71,9% of the negatives are so by "personal assesment" ) and even the results (with around 16,8% of false positives on asymptomatics) are pretty bad considering the prior probabilities of having the coronavirus or not. If you are asymptomatic and this test gives you a positive result, you are still far more likely to not have the coronavirus given the prior probabilities.

[@bluestar2253]

I always cringe when I hear people just AI'ed everything they can get their hands on!