Teachable Machine 3: Sound Classifiication

[@aditya95sriram]

I can't believe how easy and accessible all of this has become.

[@RaysOfPivot]

It's so powerful for so few images too! And it's only going to get easier and more powerful as the years pass!

[@realcygnus]

True that !

[@JohnDoe-zc4mu]

It's still a small dataset and pretty common stuff. But I agree that cloud providers are making AI at scale very simple!

[@realcygnus]

@@JohnDoe-zc4mu I was thinking more in terms of like the evolution of the entire "software stack", so to speak. Just "accessing the audio / video" for instance, in the old days, say the early 90'sish, such things took much more effort/code. You be there hours dig'n through a VERY poorly documented telephone book sized Win32 API. If you got lucky, a colleague might know some useful info or you'd find a specific book. Nor would it be so platform independent. Later, sights like stackexchange became priceless instantly. Of course the relatively recent successes of AI/ML is nothing to shake a stick at either.

[@slackerengi2401]

Information age my man You can learn nearly anything online

[@TobiasTrofast]

Oh Daniel, the people on my commute bus are probably wondering why I laugh loudly, but I feel sorry for them not knowing about your extraordinary tutorials. Thanks a thousand times !

[@abhaykashyapa]

I tried it with my flute and recognise the notes! and it worked! This helped me a lot! These Small things bring a lot of happiness!!

[@TheCodingTrain]

amazing to hear!

[@DarshanSenTheComposer]

It's so satisfying to watch these machine learning tutorials. You explain everything so well. Great way to build an intuition. Keep up the great work! :)

[@timsecond]

This is the most passionate programming channel I've ever found!! 😄😄

[@Amr-Ibrahim-AI]

This was lovely. I love how you are unlocking a lot of imagination here. Thanks Dan!

[@cyrilleb.6985]

Been watching your videos for a couple years already, keep it up 🚂 Also love to look through the descriptions for sneak peaks :)

[@sinom]

Teachable machine is really cool with how amazingly simple it is.

[@sonik382]

Thank you Dan, this is fantastic! Your channel deserves much more subs for sure!

[@UserUnknown07]

I actually saw the Google's video on this today, and wanted to know more about it and guess what I saw 🤩 . . . Loved the "Don't switch the tabs!" song... 🤩

[@anushaupadhyay7619]

Great work !! Keep posting creative videos like these. Btw u got a new subscriber

[@raphaelmorgan2307]

i love that it sometimes thinks ur voice is the ukulele ig your voice is musical

[@yuangchengyuancheng4460]

The best education video ever.

[@geoffwagner4935]

I hav to admit, i didn't even expect a tuned Ukulele, i'm slightly convinced he actually knows wut he is doing with a Ukulele

[@swatilekha_roy]

Amazing work, Dan!! I was wondering if it's possible to have two models working simultaneously or not! For instance, a pose model to detect poses while speech and an audio model to detect fumbling sounds?

[@nicholasdejong7112]

You are on FIRE!

[@g4yktzgjx6]

There is a certain kind of approximation going on when it has to classify the sound. You spamming the E-sound will create very identical samples. But maybe if you let the E sound play out for instance wait 1, 2, 3, 4, 5 seconds before playing E again will create more different samples and I believe that will help the classifier to better understand the different stages of an E-sound.

[@TheCodingTrain]

Oh, interesting point, thank you!

[@dittilio]

@@TheCodingTrain similar to above, I get the impression that the fundamental frequency may be too heavily dominating the classifier. Perhaps if the chord (or note) was played, and THEN recording starts so it learns only the decay (harmony) rather than the attack, it may have better delineation. The other option of course is to train the other way around, and combine passive background and the decaying chords you want to train, so it may weight the C-E-G attack more heavily. If I'm wrong I'd love to know, I don't code (yet) so I'm probably missing a lot of basic things.

[@geoffwagner4935]

this is all outrageous. and it's all json, you can train it, and put it in one box. nuthing here , really "has" to be cdn . even teachable machine is just making a json for us here

[@sainath66666]

Can this be used to login a user into the website using his voice? How safe can this be?

[@IceMetalPunk]

It can, but it shouldn't be. It's not at all secure to do it that way.

[@marcustulliuscicero9512]

That'd be pretty cool, something where you could record a phrase as your password and speak it to unlock.

[@nayaart6373]

very good video.... i learned alot from it .. thank you so much for your time. i have a question: what if i want to merge the sound and the video... like when there is a train picture and train voice it will give 100% its a train, but actually what i really want is a face recognition and to give info if this is an angry person, or maybe asking for help?! so mixing the video with voice that can give more accuracy ! please answer me

[@datdemlomikstik4443]

You would need 2 models - one for image and one for audio. Then you could run them both at the same time, and combine the results (labels or confidence scores) in your own code.

[@nikhilvaid6295]

i had an error of sampling rate it says "Mismatch in sampling rate: Expected: 44100; Actual: 48000 " how to resolve it ??

[@neeharikaroyal9293]

Superb 👌I am enjoying learning 🙂

[@axu6207]

Thank you for sharing this amazing resource! 🌈🌈

[@jasonzhang4584]

Great inspiration, Thank you Dan!

[@abevriens]

Black dot illusion at 2:54!

[@yoogyeongshin1031]

Sorry for commenting on past videos, I have some questions. Can Teachable Machine learn language differences? For example, is it possible to distinguish between Korean and English? It's my first time using this program, so I don't know how to use it. I wonder what to use for background noise if possible.(Sorry for my bad English)

[@ZenZed]

I tried to import this to keras but I have this error : Unable to open file (file signature not found)

[@swatilekha_roy]

Amazing tutorial!!!!

[@hikushii3233]

what do you think about challenge to code minesweper but in 3D

[@yoniziv]

Thank you so much Love the coding train!

[@شرطةمكافحةالحكاكين]

Man you know icant speak english ispeak arabic but i understand you ilove you man❤❤❤

[@gnrockyou]

So where do I start looking up for doing a song classificator? I would love to build an interaction of the songs I play on the piano with a sketch in p5.js.

[@learnelectronicswithdevans5422]

I am not seen TensorFlow lite option for speech audio, How can I run this model on mobile or RaspberryPi

[@petaca39]

hello good explanation a query can generate a PWA (progressive web application), using ml5 and p5.js. If it is possible to do it, you would have a link with information on how to achieve it, it would be enough with this same example of the video but in PWA format.

[@slobodannesic9175]

Could this identify different type of vehicles?

[@vivisaens3487]

Thanks for sharing this video! I have a question, I'm trainig with classical music, I created 3 classes (Liszt, Beethoven, Chopin) Each class contains 100 audio sample I clicked training 2 times and then, after "model trained" appeared , I started to play chopin. But percentages related to each class are constantly in movement, the first higher percentage was related to Liszt and then Chopin, but which one I have to consider? I have to choose just one value? Thank you In advance

[@TheCodingTrain]

Our Coding Train Discord is a great place to get help with coding questions ! discord.gg/hPuGy2g - The Coding Train Team

[@vivisaens3487]

@@TheCodingTrain thanks so much!!! I'll try it ^^

[@RustuYucel]

Wondering if this can be used to voice recognition? Any help on this?

[@JohnAmose]

Can prerecorded audio be used ?

[@charlotteloke4825]

Hi, why after inserting the let emoji = "" .... (part) my screen can't display anything?

[@francisjosephlarita2818]

Is it possible to transfer the program into the arduino?

[@mamupelu565]

Can I use this to scan a paper then make it recognize words there?

[@hamzahassan8024]

I love thw tutorials

[@4541047]

Do you think it will be accurate enough with distinguish other people voices??? I'd like to try it with my different friends voice model training

[@TheCodingTrain]

Give it a try!

[@mordiaky]

Oh man, you need to teach it brian waves like up down left right just by your thoughts. Now all you need is some device to record your brain waves.

[@jimbarchuk]

Google DIY EEG.

[@threeMetreJim]

Isn't that what TESLA is doing?

[@Sposhal]

hi sir, how to use the model in android studio?

[@aminelouguit2008]

why we can not use a zip file instide of mic ?

can i use teacheble machine for my own project/product?

[@TheCodingTrain]

Yes, absolutely!

[@akshaybhagwat2270]

I am doing image train project on teachable machine.... I have added 3 class like class-A , Class-B and Class-C....if no any img can show from A,B,C but they show results for that unknown img in A,B,C.... How to train model if nothing then show nothing...I have added class for nothing also but they show A,B,C class results

[@zingnolan]

Love you always

[@anlpereira]

Hi, do you have this code in Python? Thanks

[@kawaiifreak97]

someone please sample "dont switch the tabs" 5:54

[@clipi_]

Yay!!

[@kunion2512]

you are humor ! i like it

[@z4zuse]

A more accurate background sound would have been you talking :)

[@chitaozinhodoextrativismo4108]

Hello shifftman. I challenge you to replicate the marangoni effect in JavaScript

[@maheshm4292]

Sir, how can I use this in just JavaScript??? With out p5.js???

[@TheCodingTrain]

Yes! Take a look at the plain JavaScript tab in the code snippets at Teachable Machine.

[@maheshm4292]

@@TheCodingTrain alright sir.

[@TioMuhamadNur]

What’s software name did you use? Actually, i’m beginner

[@TheCodingTrain]

I am using editor.p5js.org/ and teachablemachine.withgoogle.com/ and ml5js.org/ for this project!

[@bo3oof]

thank you

[@rfyorfyo5022]

isn't a python version for this?

[@enivaldobonelli]

Great!

[@michellenikeetha7453]

who has any more ideas for more projects to do using this teachable machine: audio?

[@Ninjacowxtpj]

ILY

[@secretmystery8305]

Nice

[@prachetass]

Request : Please do bubble shooter or match 3 game

[@isaacchen3857]

Cool idea! I wonder what sort of data structure would be best to keep track of the bubbles?

[@prachetass]

@@isaacchen3857 i started writing the code and i am stuck ..may be i messed something up and now can't keep track ...if you are interested I will send the link... It's not all my code and no helpful comments are der,. :P

[@isaacchen3857]

@@prachetass sorry for the late response. If you're still working on it, I'd love to help.

[@prachetass]

I am new to coding ...the code looks bad ....here it is editor.p5js.org/prachetas/sketches/vuZJWRqSW

[@isaacchen3857]

@@prachetass That's awesome! I'm at uni right now and finals are coming up, but I'll take a look when I get the chance. Have a great day :)

[@장성숙-l9z]

great wow

[@ilayws4448]

I used it to check if my eyes are more similar to my mother's or my father's :D

[@SpaceCoding]

Meoooww

[@JazzieGames]

WHY Unlisted ?

[@geraldbustos]

Unlisted

[@iminni3459]

I will be public today or tomorrow probably, it's so that the videos don't all come out at the same time.

[@prachetass]

Here: editor.p5js.org/prachetas/sketches/zI_nc0C1d

[@spyrosmiltiadous5229]

MEOOWWWW

[@davidpanic]

meow

[@threeMetreJim]

So you make a great model for something, and Google then has it too, to make money from your work. Everybody and his dog knows Google, so you'd have no chance, and probably face some sort of infringement claim too. Questionable ethics, for sure.