Placeholder Image

Subtitles section Play video

  • Okay.

  • I think you should hear me now.

  • How do you don't see me?

  • But you hear me?

  • I have a new button that utes and unused my microphone.

  • And in a moment, I have a button.

  • It's which is to the camera that I'm in on.

  • This song ends in about 14 seconds.

  • 13.

  • Well, I'm counting.

  • Waitress flows.

  • 76543 too.

  • And you're on.

  • Do you see me?

  • Do you hear me?

  • Is it Monday?

  • No, Tuesday.

  • I don't know what day it is.

  • I do.

  • It's Tuesday.

  • Welcome to wth e coding train.

  • I feel like I'm speaking very loud and people in the hallway are going to listen and hear me, and I feel embarrassed.

  • Uh huh.

  • So this is unusual that I'm here on a Tuesday.

  • I missed Friday.

  • I'm having a hard time.

  • I'm being totally honest.

  • Here's the good news.

  • I have I've set up.

  • I've worked very hard with a lot of people's health.

  • I can barely take any credit for this, but I have a new studio that I am live broadcasting from, and soon I'm hoping to make this studio available for other people in particular to the N Y U community because this is a room that's at you where I broadcast my other separate thing, independent thing, the coding trade.

  • It's very complicated.

  • So studio is going well.

  • I have new buttons.

  • I commute my microphone.

  • Watch this.

  • I can I have this button that makes my laptop disappear.

  • But, look, isn't that weird?

  • Look at that.

  • That is crazy.

  • But so if I need to, like, type in a secret password, that okay and then I could make myself disappear.

  • They're going now if I just need to show you the screen, there we go.

  • I feel like the focus might be off a little bit.

  • I don't I need I am gonna get I'm going to get someone to come help me with tuning the cameras a bit focused white balance, that sort of thing.

  • I know.

  • Also, these lights are reflecting in my glasses.

  • I turned them a little bit, so maybe that's a little bit better.

  • I don't know.

  • I do, but so things are good.

  • That's what I was saying.

  • Things are good in that the studio seems to be working, um, all that sort of stuff, But I'm having trouble getting sort of regular schedule going.

  • So today was like a target of opportunity.

  • I have a round till 11.

  • 30 little bit past 11.

  • 30 free.

  • So I thought I would try to do a live stream I missed last week.

  • So I'm gonna try to do to this week.

  • My goal is once a week.

  • Um, but one of the reasons why I've actually had a little bit of trouble Livestreaming finding time to live stream this.

  • This may fill you with joy or might disappoint you.

  • I have no idea.

  • But I have a new video that hopefully will be out later this week.

  • That that that I did a coding challenge in that was not from this studio in its fact from a different secret location that I will not say anything more about when the video comes out.

  • You will see this new secret location.

  • I don't have the capability of life streaming from there.

  • So it was recorded and then edited.

  • But I hope to be doing these new this new style of video in addition to the Lifestream not replacing, but in addition and we'll see how that goes.

  • So I look forward to when that comes out later this week.

  • I don't being very vague about it.

  • I look forward to hearing your feedback.

  • I will provide a sneak preview to this video in our coding trained member and patron slack group.

  • So if you're thinking about wanting to join and you want to see a video like 24 hours earlier than somebody else, I don't know why.

  • That's a thing.

  • Um, now would be the time to do so.

  • I'm also planning Thio.

  • I'm looking a little on the bright side, says Alka, but I'm looking sharp.

  • Thank you.

  • Um, the problem is, when I look at the chat, I lose my train.

  • I thought it's a sad life.

  • It's so sad.

  • I will have some new announcements to about the rewards program for mem coating trained members.

  • So I'm revamping that got some new ideas that are kicking around.

  • Some new stuff will be launching.

  • So hopefully you hear about that later this week or next week.

  • Um, and just on cue, we have a new member.

  • A wren, be so welcome.

  • A randy now.

  • Uh oh, no, I haven't implemented this yet.

  • Soon.

  • There.

  • There's some new new tools in the YouTube membership thing.

  • I shouldn't go on too long about this because most of you are just you're watching, and you don't wanna hear me talk about the membership stuff.

  • But once you've joined, there should be information about how to get your slack inviting all that sort of stuff.

  • It might not have shown up for you, but we're fixing up that stuff in hopefully by the end of this week, that will be in better shape.

  • Um, okay.

  • Huh.

  • So let me just first before I, um, go off into my rambling nonsense.

  • My goal for today is to do to tutorials on two functions, one to trail on each function.

  • There's two functions.

  • One tutorial for each that makes two in the Ml five Jazz Library.

  • If you're not familiar with Ml five, Jess, now's a good time for me to open up my Web browser on dhe.

  • Tell you the ml five.

  • Jesse's a friendly machine learning library for the web.

  • It is a library that is built on top of tensorflow Js.

  • It is one that I work on with a lot of different collaborators here.

  • And will you?

  • And throughout the greater open source community.

  • And so I have a playlist going of video tutorials.

  • There's some really exciting what I find to be incredibly exciting new features that are in development and amplify.

  • They're not part of the release yet.

  • If I can get my act together later this week, I'll do a second last dream with those new features.

  • But I wanted to fill in some gaps about a couple models that air here in Ml five that I have not touched on in particular You net and body picks.

  • And both of those are examples of image segmentation.

  • So high code and garden with CJ Um I'm ah, I got another new member.

  • Thing is great.

  • I love this member thing.

  • I'm so happy I'm gonna get some lights that blink or something.

  • When somebody joins, That'll be fun.

  • Uh, shriek, Shaikh, Thank you so much when you first joined you get a little this dot icon, What is the world?

  • What is this world?

  • And my boss?

  • Is it okay that I'm buying into this?

  • That you'd like to sign up for this membership thing and you get emoji icon?

  • Okay, Um so, uh 00 the website starting to look different already.

  • We're doing some new stuff with the website.

  • I'm super excited about it, but, um so this is that the goal?

  • That's really my only goal for today.

  • I want to show community contributions I want to do to tutorials.

  • I want to thank the sponsor of the coding trade watches.

  • I even have a button on my no sponsor Delgado stream deck that will bring this come up.

  • So thank you.

  • I have the Lifestream is once again sponsored.

  • I took a break from that for a little while.

  • It's figuring things out, but I think I'm settled.

  • So thank you, Thio Brilliant dot or ge Um, if you're interested in finding out more about brilliant is that right?

  • Is it that I Hopefully I opened up the correct banner?

  • I can't remember with the or just coding train.

  • Hold on.

  • I might have to fix Miss, This is welcome.

  • Thio, My incompetency.

  • Hold on.

  • I know how to check this.

  • Just talk amongst yourself for a second.

  • And remember, be a coating Lies dream not just to thank the new members live stream.

  • We're like, this is very hard to use this computer over here.

  • PC, Can anybody see the description of this video?

  • Kids, wait.

  • I know where to look.

  • I have it.

  • Definitely having correct in the description.

  • That's where I have to look, I'll swap it out.

  • Here's the Lifestream.

  • It's me.

  • Oh, yeah, that's a mistake.

  • When we fix that banner live exciting.

  • I thought I was being so good and I was gonna have no issues.

  • I don't have it on the worst.

  • Well, I think we're going.

  • All right.

  • Well, I will take a break.

  • I'll take a break and fix that later.

  • All right, back roads.

  • So if you're interested in learning more about brilliant coat of brilliant dot org's slash coding train not Nova Nova.

  • I mean, the is part of the coding train, but this particular year Ellis, just last tree.

  • Want, want, want, want.

  • Okay, um, let me.

  • Okay.

  • I mean, next thing in the next segment I want to do is show you community contributions, But I have something else I have to show you first.

  • It's very warm in here.

  • Definitely gonna have to send an email about the air.

  • I could feel myself palpitating new member again.

  • This is crazy.

  • Hello, Jonathan Barbeau.

  • This is nuts.

  • I've never had so many new members all at once.

  • So I'm about to change to a different camera view.

  • Here we go to press this button over.

  • Now it's This camera is set with a lot of automatic settings.

  • So I think it's going to start to look right up because it's like auto white balancing.

  • Um and it also has auto focus.

  • Hello.

  • But I am very excited to show you something.

  • This isn't a smart surface, its magnetic, so I can hang things up here which I'm very excited about.

  • That should open up a lot of possibilities.

  • And this is, ah, a sheet of draw a dry erase sheet.

  • So I can now right here.

  • And once again, I have my ability to diagram.

  • I don't know how visible that is for you, and certainly as I come closer, it's gonna go out of focus, so I probably should focus the camera.

  • I'm specifically just on the wall on dhe tune, the white balance and all that stuff.

  • But at least for now, have this crease green cloth of At least for now.

  • I can use this as a white board.

  • So that's what I'll be doing today.

  • You might remember last week.

  • It's over here.

  • You can't really see it because I could think was my white board from last week, which still has all the stuff that I listed on it.

  • But it's just over here in the corner right now.

  • The thing that's great about this is it seemed far as I can tell.

  • It's Matt, and it's not really reflecting the lights to too much.

  • So it should.

  • This should work better than what I had before.

  • All right, so that's the other new thing.

  • And then I come over here and press this button and then I am back over here.

  • Uh, okay.

  • People are asking.

  • When was the last time there was a live stream?

  • With Simon absent, you would think something must be wrong.

  • There's Simon's, not here.

  • Oh, and I don't have the select channel open.

  • Let me get that open.

  • Here we go.

  • Um ah.

  • Oh, my God.

  • Too many things.

  • How do we do this?

  • There we go.

  • Very faint.

  • Um, Okay.

  • Separate.

  • All right.

  • So we'll see how this goes today.

  • Each day I make each time my lifestream.

  • I make a small, incremental improvement.

  • I found this bug in the kitchen, by the way.

  • I love it.

  • I think this is a Scrabble mug.

  • It's gotta end with one and then all the other letter distributions back here.

  • Okay, Now, if you recall one of the things that happens on the coding train is I make a coating example like a coating challenge.

  • And last week previously previously on Decoding Train I programmed ukulele tuner.

  • Now I should mention and many of you rightly pointed this out.

  • Mine don't have any of you who are actually watching this live with the regular people who showed up with one video that I made.

  • Many rightly pointed out that this is not a particularly accurate or efficient way to make a ukulele tuner or to accurately estimate pitch based on audio input.

  • And they're certainly like highly specific mathematical ways of doing that.

  • Related to FFT algorithms and sound analysis.

  • Maur, what I was attempting to demonstrate was a particular machine learning pitch detection model that happens to be an ml five's.

  • It was a nice opportunity to use it to create this tuner.

  • So if I click here onto the challenge.

  • We could see the videos here, and I'm gonna just quickly run mine, see if it still works.

  • It's not very promising.

  • Oh, I know what the issue is.

  • How This is a weird This is Ah, you know what?

  • This is a good thing to bring up.

  • So why is this not working?

  • I ask you the viewing audience, and then you will near this question about 30 seconds from now.

  • And I won't be able to wait till you try to suggest an answer.

  • So the reason why this is not working is chrome.

  • Uh, for good reason, I would say, but I think this is a debatable point, but will not begin any audio source any playing of audio or recording of audio.

  • That's probably even more important in terms of privacy without a user interaction first.

  • So in other words, we want the crow.

  • I suppose the reasoning is to protect against a website that might surreptitiously record your audio.

  • You can imagine how problematic that could be.

  • Um, so I didn't so in this eye problem for me to actually have programmed this ukulele tune or so that it would work we need a button, like start listening or something, which I didn't include, because if I just go to the actual editor view, there is a user interaction that happens on this page.

  • Me pressing this run button on.

  • Now it works.

  • So while I'm in the editor of you, it happens to work.

  • But when I'm not, it doesn't.

  • So I should be able to tune his ukulele.

  • You do, Do you?

  • I thought that was this was the project I made.

  • I would say the big issues with this, the things that I don't like about it are mostly all have to do with user inter interaction and visual design.

  • And I would say these are not my strength.

  • I don't know what my strengths are, but they're certainly not that.

  • And so what I enjoy doing with these cutting challenge is is making this sort of, like technical demonstration of the idea and asking the viewing audience to make their own version of it.

  • Maybe with a creative twist or being more thoughtful about the interaction design.

  • Lots of possibilities there.

  • So, um, I'm gonna close my version out, and I'm going to go now to this page, and I'm sure you'll like so many of these came in.

  • Okay, well, three thumb are admittedly, three of them are from Simon of, but there's a lot of these contributions.

  • So I would like to spend more time in my life streams sharing work that people are making from the audience and from elsewhere in the world.

  • You know, I really want a future.

  • A lot of different kinds of people who are doing stuff with code.

  • Um um, from And so I'm gonna take a moment to do that right now.

  • So I'm just gonna click through these in order and we'll sort of see how they work and give my commentary on them.

  • What could go wrong?

  • What could possibly go wrong?

  • So let's try this one by Luciano.

  • Tony.

  • Uh, wow.

  • I don't need to translate the page.

  • Okay, so let's see here.

  • I think way.

  • So I think we're gonna run into this issue.

  • We can look to see the audio.

  • So this is what I'm talking about.

  • User gesture needed to start audio context.

  • Please click.

  • So I don't know how to force this toe work.

  • Like if I can click fast enough Um oh, there we go.

  • So I got lucky, so I just clicked a bunch of times.

  • So now let's try this.

  • So I like this.

  • I like that so that one thing I like about this is it's closer to the what you might typically see with the tuner where there's kind of a dial, and the DIA list when it's centered, is at the right frequency.

  • I might say that this is jumping around a lot.

  • A lot of that has to do with the pitch detection huddle on the fact that I'm speaking and there's noise in this room.

  • But in some type of interpolation, Um, some other visual indicator right now I think this is a problem with mine is just like on or off quality of Is it tuned or not?

  • You want to see my getting closer or further, I think could help here.

  • But I like this idea of having it.

  • B'more dial like Okay, let's try this one.

  • My mind, Bill, I think I'm just gonna click into the editor.

  • I love this.

  • I'm like filling up a circle.

  • Let's try changing the note and see what happens.

  • You can see how the circle's getting smaller again.

  • We've got this a lot of noise and this sort of green or red, no interpolation there, which I think could could potentially help.

  • I like that.

  • There's this option of changing it to a guitar that is cool.

  • So the notes are different.

  • What it's looking for.

  • Basically the tire, sometimes A C J.

  • Writes Sometimes having the death tools open before the page loads allows the mike to work without interaction on.

  • I see Simon Tigers is typing, so I think Simon has made it to the Livestream.

  • Everybody do not.

  • Do not worry, Everything's OK in the world.

  • Okay, um, next we have the ukulele tuner with a pitch trail by original ing.

  • Let's try this one.

  • Oh, whoa, Oh, I love this.

  • It's almost like those you know, Guitar Hero style games were The notes are coming at you and you have to time the interaction.

  • There's this, and I love the idea that I could create a picture from playing music here.

  • In a way, I wonder if there's something to this beyond tuning, which is a way of visualizing music and sound wonderful idea here.

  • Ukulele tuner, same model new design by David Snyder.

  • Oh, whoa.

  • Look at this.

  • Now, this is great.

  • Embedding the tone generator is a very smart idea on then.

  • Here we've really got a nice dial going.

  • Let's play.

  • So let's move that this around.

  • Something seems a little off with the position of the dial again.

  • This is not the perfect, um, the perfect way of doing pitch detection.

  • But I this Now we're really starting to see something that has a very traditional dialect design.

  • I like that.

  • There's a little indicator of what the mike is doing.

  • Some how this worked right out of the box.

  • I wonder if if you know why.

  • If there is some kind of like user interaction embedded that I don't realize that I've done here, but this is great to see.

  • What if I play different Notes?

  • It's auto.

  • Detecting the note.

  • Great.

  • Great job.

  • Thank you.

  • David.

  • Um, and now Whoops.

  • We have three submissions from Simon.

  • Let's look at them tuning all strings at once.

  • What is that?

  • Even pose that possible?

  • Let's see if the dep Uh huh.

  • Yeah.

  • So you think that deaf tools opening would help, But I'm just going to go to here and I'm gonna do this.

  • I like this.

  • So let's strike to end a little bit.

  • I like the yellow and I like, sort of like yellow color as a kind of warning sign.

  • I'm almost there.

  • Everything works pretty well.

  • Having trouble getting my years and so great, Wonderful.

  • I I like this of having all of them on one page here.

  • That's a nice design idea.

  • Click to tune the next drink.

  • So I think this or like different interaction design ideas right where one mind just auto to text.

  • Which string and other possibility is showing all of the strings.

  • And here's one where it's actually giving me instructions.

  • So this is nice.

  • It's like a tutorial.

  • Walk through is always helpful to the user G.

  • Is this one on?

  • I suppose if I click, it signed me to do C Street, so I like this as a design idea as well.

  • And then random number generator this.

  • I don't know what this is.

  • Let's see.

  • All right.

  • I'm fascinated by this.

  • What is going on here, Simon?

  • Temp air pitch got pitch.

  • Is it generating a random number based on the audio that's coming in when there's four submissions.

  • Violent rights.

  • I feel like this kind of project would be much easier in Max MSP.

  • Yes, O P five Js.

  • The browser is not necessarily the optimal context for doing for creating this kind of sound processing applications.

  • Certainly other programming environments are geared, and you're entirely around working with audio, so I think that's a good point.

  • Um okay, so I don't, um Simon, I love to hear more about this.

  • Random number generator 123 only.

  • See, three, according to Simon.

  • Therefore await original ukulele tune, right?

  • Somehow I missed that one.

  • Okay, so this is just a version of it That's much like mine with a slightly different interface.

  • Okay.

  • Uh, great.

  • So thank you, everyone for your ukulele tuner Submissions.

  • I really enjoyed those.

  • That next segment next segment, I'm gonna look att some or community contributions.

  • So let me go, Thio.

  • You know me.

  • And I'm obsessed with this idea of spinning a wheel.

  • Um, so let me go to get hub dot com slash coding train slash website and we'll see David Viewer David has created a wheel that actually will randomly pick a viewer contribution to feature.

  • So let's This is a little trial run here of this idea.

  • Maybe we're gonna work on this more so I think this is I actually don't know.

  • I think this is pulling from a spreadsheet of, uh, user.

  • I don't like the word user.

  • You know what the word is, and this is not my idea.

  • I'm sort of pulling this from a talk.

  • Our presentation that Dan Pfeiffer, who is a wonderful programmer and activists and person on the Internet, people there, people, the people that users they're not users to manipulate and have make them use your stuff.

  • They're just people.

  • So this is the wheel that its numbers.

  • So I've reduced people to numbers just maybe a little bit reductive at the word.

  • But these are all tied to a particular of per person viewer contribution, and each time it's going to pick 10 random ones.

  • And once it picks 10 random ones, I can now spin the wheel on.

  • Dhe may be nice to see what the people's names are.

  • I can also do my little drum roll here.

  • I don't know what this one is.

  • Number four.

  • Who is number four passengers.

  • Thank you.

  • C J.

  • Passengers on the train.

  • You know, the train thing was just like I needed a name for this channel within, like, 24 hours.

  • It's like somebody coated crab.

  • Like I like that.

  • But I don't know that the idea of the train theme has really come all from hell for me.

  • Really?

  • This is really a channel about coding and rainbows.

  • I'm being perfectly honest.

  • Just wanna put that out there just for a minute, would put that out there.

  • Okay, back out to this.

  • So I believe if I my mug, is, um, blocking this.

  • But I believe now if I click this button remove and view selected, it's going to take me to this.

  • So we picked.

  • Here we go.

  • I don't Number four.

  • This'll challenge.

  • This one is based on is Perlin noise loops.

  • And this was contributed by Maxie Monakhov.

  • All right, over.

  • Use this.

  • Now I'm gonna click on it under view it here.

  • You can open a new window.

  • It did even.

  • Oh, cool.

  • Wow.

  • So I believe this challenge, if I remember correctly, was showing how to close a loop all the way around a circular path with the kind of the radius and any given angle being tied to Pearl in noise on dhe.

  • What's a nice application of that?

  • It looks like making, like, wobbly little wiggly asteroids.

  • I think he's in asteroids game.

  • This is really good.

  • Look out.

  • Look how elaborate this is.

  • So I'm gonna make this easy.

  • I'm gonna give myself three lives, and I'm gonna start the more difficulty level the smaller number of lives, the more points you get controlled by arrows.

  • Press escape to return to the menu.

  • Okay, we go.

  • Wow, this is great.

  • I'm sorry.

  • I have to end my life stream now and spend the rest of the day playing this.

  • Wouldn't it be fun to work on ah, body in a I a little evolutionary algorithm to evolve a strategy for playing this game.

  • Wonderful.

  • Wonderful job.

  • Thank you.

  • Two already forgot who this was.

  • Maxim Monakhov for this variation on pearly noise lives.

  • Let's look at one more.

  • Let's look at one more Just I want to be conscientious about the time.

  • I do have some tutorials I want to do today.

  • Um, so let's look at one more.

  • Uh, spin the wheel number three.

  • This original challenges blobby Well, there's going to feel like you have a lot of challenges based on blob like shapes and its continued by Grilli 86.

  • Let's take a look and see Well, there we go.

  • So this is nice.

  • So this looks like a pretty exact replica of my challenge.

  • If I remember correctly, I don't think I had this sort of changing color.

  • So it looks like it's been added.

  • Here is that the color is changing according to maybe a noise algorithm, but also, it looks like if I oh, no, there's you know what's nice about this is seeing it on code pen and actually looking at the code looks like there's a bunch of other parameters that air changing and happening there's like lots of war going on.

  • So this is really nice.

  • I mean, this is actually oddly, one of my favorite things to visualize is just the kind of undulating, gelatinous like shape, things that are spongy, that are soft or quite hard to do with code and very satisfying, I think, to play around with and and use Okay, great job that ends that segments have segments now I think it's going very well, right?

  • Things going well?

  • How's it going?

  • It's going very well.

  • Thank you.

  • No.

  • All right.

  • I'm gonna close these windows out, people, I Mm.

  • I am a smart surface.

  • I am a smart surface.

  • People like me.

  • Uh, just, you know, daily affirmation.

  • Nothing's wrong with little daily affirmation.

  • All right, back over here.

  • So if you're wondering how to submit your own coating trained community version viewer person submitted passenger very ancient on a challenge goto the website.

  • And this is a link to add your own version.

  • And it's a little bit tricky.

  • You have thio use the get hub website.

  • You have to you do something called a pole request.

  • But we are here.

  • We, the coding trade community, are here to be friendly and welcoming and help you through that process.

  • You can't do it wrong.

  • You can't break anything.

  • And even if you did break something, that would be awesome.

  • If you can figure out how to break the website, that would be fun and a learning experience for all of us.

  • So don't be shy.

  • If you think that your contribution isn't valuable, you're wrong.

  • It is valuable.

  • Please submit it.

  • I even have a video.

  • I think that goes through this, like community contribution coding train.

  • I just Google that There we go.

  • How to add your contribution to the coding train website.

  • So this video hopefully give you a little more context?

  • All right.

  • Uh, did I plug my cat instagram yet today?

  • All right, So here's what.

  • Here's what I've got a little over an hour and I'm three more things I am going to do a quick tutorial on.

  • What am I saying?

  • Um oh, I'm gonna do I'm gonna look at the you Net model in ml five.

  • Let's do a daily challenge on brilliant dot or ge, and then go look at body picks.

  • An amplifier.

  • That's what's happening for the rest of this live stream today.

  • So give me a second here to get set up for the next bit.

  • All right?

  • By the way, I've been seeing this warning slash error a lot recently, and I don't know what it is.

  • A cookie associated with a cross site resource at cloud fair dot com was set without the same site attributes.

  • Why is this appearing?

  • Is have something to do with loading the P five library from the CD en Oh, it's this must be right, because that's cloud fair.

  • No more warning, huh?

  • So is this.

  • What is this?

  • I don't wantto get off track here, but I'm wondering, is this an issue with Cloud Fair with the P five Web editor?

  • Who knows?

  • Um, right.

  • Somebody can write to me, or I probably might not see it in the chat right now, but we moved in.

  • Somebody can help me with this.

  • It's like I see Alka is typing, and I'm almost certainly certainly answers about to come.

  • Um, but this is not a new era that I'm super familiar with, So I want to do this tutorial, not in the p five Web editor, because I'm not super familiar.

  • I haven't, Actually.

  • This is a thing that I do, which is?

  • I make tutorial.

  • That stuff I haven't used just right off the cuff.

  • Why not?

  • Um but I'm not super familiar.

  • I've really done a lot of work with this you net function, so I think debugging it or figuring out what's going on might be a little easier in the browser.

  • You can close the what's new part of the consul.

  • I did that.

  • I'm not so sure about the cookie area.

  • Okay.

  • Um all right.

  • So one thing I can do it at a minimum is I can tell it to not show my warnings on, so I'm gonna just take this off and then we won't see the warnings.

  • All right, So what do I need?

  • I want a couple things before I start this tutorial.

  • I want the ml five website close The web editor.

  • Um, I want Thio reference, so I'm gonna spell this wrong, but I'm hoping Google will help me.

  • Um XIII ds get hub and Twitter.

  • Um, and website Ali, FBI.

  • Maybe somebody could help me.

  • Who's watching?

  • I should have asked.

  • AII this am I pronouncing this within the realm of reasonable nous for Ah, silly American.

  • And I am Zaida out out all your fi XIII dahlia fi and somebody help me with this.

  • So while I'm waiting for maybe the chat to help me with pronunciation, Adam asked Unit.

  • Are you going to do some segmentation?

  • Yes.

  • So I am here to demonstrate how to use a pre trained image segmentation model which has the unit architecture in the Ml five Jazz library.

  • So this is not a tutorial on the unit architecture coating it from scratch with tensorflow and C plus.

  • Plus, I'm really just using a very high, high level tool to have a pre train model and sort of look at image segmentation in in ml five.

  • Um, Okay.

  • Okay.

  • You see if I can, uh Well, thank you.

  • I got some sort of super chat thing from Christopher Jessup.

  • Thank you to Whoa for a lot of money.

  • Thank you, Christopher.

  • Doesn't this make me feel very awkward?

  • Not comfortable.

  • But I get it.

  • I appreciate it up.

  • Robo train with little Oh, that's very kind of you.

  • Very much appreciated.

  • It's more like Zaid.

  • So did I get that right, armer Zaid?

  • Um yes, people are, um all right, do coding challenges.

  • Holy just up reading the chat.

  • I was looking for something here.

  • I was gonna at make some people moderators, but Okay, um, I didn't notice.

  • It's just that it takes me take some time, right?

  • What?

  • You don't understand.

  • I mean, sometimes I don't notice, but you just really have to be aware.

  • There is quite a significant delay in what you're seeing me versus the real time of the chat.

  • So anything that I respond to in the chat Hap you won't see it for at least 30 seconds.

  • Um, all right.

  • Uh, okay.

  • Zaidi.

  • Okay.

  • Okay.

  • Here we go.

  • I'm like, Do I need to cycle the cameras?

  • Are they gonna shut off?

  • It's 10.

  • 30.

  • I've got an hour here, so I think we're in good shape on DS here.

  • All right.

  • Hello, and welcome to another tutorial.

  • Part of the Beginner's Guide to Machine Learning in Java script with Ml five.

  • Jess.

  • So, in today's tutorial, I want to talk about a particular I've seen.

  • You could hear that.

  • Yes.

  • Uh, we can hear that.

  • Yes.

  • Uh huh.

  • I'll just start over.

  • But I almost feel like I want to keep that sort of funny.

  • Um, hello and welcome to another video is part of the Beginner's Guide Thio machine learning in job script with Ml five gs.

  • So in today's in this particular video, I want to look at something called image segmentation.

  • I have a white for just looking at how well you can see that.

  • So, what do I mean by image segmentation.

  • So right now you're looking at an image of me, and there are different parts of the image that we, as human beings, could dissect or deconstruct.

  • So let me try to do a very poor job of drawing a version of this myself on Baby.

  • There's also sort of this so I could think of this image is having different segments, different parts.

  • There's my head, There's my body.

  • There is the background, which is maybe a wall.

  • And then there's actually a sort of secondary wall over here.

  • That's the wall to my right.

  • And this is the wall to my left.

  • I could even think of that as another segment.

  • So in this case, I have full.

  • I'm just decided.

  • This video has four segments A, B, C and D.

  • And in a way, this is ultimately ah classifications problem.

  • What I'm doing is I'm looking at every single one of these pixels in this image and trying to classify it as one of four possible labels, and you could imagine how this could be useful.

  • Maybe you want a computer vision, Pete software to be able to find all the pixels that are cars on the road that are flowers in the garden.

  • There's so many possibilities here, and so this is there.

  • There's so many possibilities here, and there are a couple function there there.

  • What am I trying to say here?

  • Looks like there's 1/3 segment.

  • What's going on?

  • There are a lot of possible applications of this.

  • So the there's a much longer story toe how you might collect a data set that is labeled and segmented train a model to recognize image segmentation.

  • And maybe this is something that I can do a deeper dive into in future videos.

  • But in this video, I want to start with a particular segmentation model that is trained on faces.

  • Um, is that what it iss?

  • Yeah, I think it's face and really just has trained on faces and really just differentiates the image into one of two categories.

  • It's the foreground of the background.

  • It's the human face, or it's not part of the human face.

  • And this is a quick way to do masking or other types of things that you might want to do in an interactive media application.

  • The architecture of this particular model is something called you Net and you net.

  • I will include a link, Thio.

  • I will include links to more.

  • Resource is about what the unit architecture is if you want to get lower level with this.

  • But this is a particular machine learning architecture that was designed and invented for a medical imaging.

  • So you could imagine how image segmentation could be useful with medical images.

  • If you're trying to pull out different parts of the anatomy or recognize certain aspects of the human body, what else do I want to say about this?

  • Um, wait, There's another super chat.

  • This is crazy Christopher Jessup.

  • Oh, it just has it twice.

  • Hopefully, that's not a mistake.

  • Thank you for the stop.

  • Um, turning like bright red here.

  • All right.

  • Okay, um, if you're looking for data sets that are available to the public with training data for image segmentation, a very well known one is called coco or common objects in context.

  • But what's the How does that make cocoa common objects in context?

  • Wouldn't that be see?

  • Oh, I see.

  • Timewise.

  • Where?

  • Okay, uh, common objects in context.

  • Oh, context is C o I got it.

  • I figured it out the cocoa data said our common objects in context.

  • Um, can you see this?

  • It gets really bright as I get closer here.

  • The white balance is a problem.

  • Um, so in the next video, I'm actually going to look at a particular image segmentation model called Body picks.

  • Which you can start to think about is how that must have some similarity with this face.

  • You net model.

  • And this particular model was trained off of the cocoa data set.

  • But let me talk to you about the unit model.

  • That's in ml five with one pre train model that's there and how it was trained and who trained it and who contributed this to ml five.

  • Uh, okay.

  • Christopher says it's not a mistake.

  • Thank you.

  • Um, but and I I appreciate that It's the last one for today.

  • Okay, Um, all right.

  • So hold on.

  • I just want Thio.

  • The pre train model, The pre trained unit model that's in ml five for you to use was trained by a researcher named Zaida off Ali.

  • If I may say that one more time because I feel very awkward about it, the pre train model that's available for you to use an ML five was trained by a researcher named Zaius Sayyed Ali.

  • If I and you can find out more about Tiede Sayyed, the the Restrain, the pre train model that's in Ml five that you can experiment with was trained by researchers.

  • I ead al you find, and you can find out more about him from Twitter Page, Get Hub and also website now include links to all those in the video's description.

  • I do want to point out here that this particular model was trained from this data set and I haven't done.

  • I haven't looked in depth at this particular data set.

  • It is a face head segmentation data set.

  • But if you look at these images, you might notice something sort of strange about them.

  • I believe that this is a data set that was created through three D simulation, and I you know, maybe some of you enterprising viewers could look into this Maur and provide some commentary in the comments that could come back in a future video and look at this data set a bit more.

  • But there's always a lot of complicated issues around collecting data of people's faces I'm making sure that you have a representative sample.

  • Are you collecting those faces with people's permission?

  • In general, I might say, if you're collecting people's faces, maybe you just shouldn't be doing that.

  • But this is one approach to doing so by I'm looking at the chat.

  • And the super chat is going off the rails.

  • But But this is This is one technique of doing it.

  • Because also, even if you get riel images, you're going to have a hand label them somehow.

  • But if you're simulating the faces and then creating the segmentation maps, this is the training data.

  • So this was the training data that Zaid used to train that model with the you Net architecture.

  • Let me let me offer a quick few.

  • Thank you's here.

  • Um, those were all Christopher shit.

  • Thank you.

  • Thank you, thank you.

  • Pop up.

  • I am your train whistle blowing person here.

  • And then also thank you to Enrique Lopez.

  • Thank you so much.

  • Very kind of you.

  • Um, this really allows me.

  • The support really allows me to have allocate more time to doing this, to bring more people into the community.

  • And I hope that I'm doing everything in the best way that I can.

  • All right, Um, now, let's look at the reference for this unit model in ml five and try to build a quick example that does the actual image segmentation.

  • Remember, we're not training the model.

  • We're not creating our own image segmentation model, which is making use of one that somebody else trained and applying it in the browser with the ml five Jess Library.

  • So first, the first thing I'll do is I'll go here on two reference and I'm going to look for, um, sorry.

  • I'm going to look for a unit right here.

  • Um, and wealth Amazing.

  • First of all, the website is totally like inactive progress of changing and being updated.

  • Wait, where am I even even on the website?

  • Because I went to hear reference Oh, yeah.

  • It's taking me two kids up that I go crazy, like I think Joey Lee, who is one of the lead maintainers of ML five J s, has been working on the website.

  • Quite it looks completely different than what I used to which I think it's fine.

  • This is making it fun.

  • Um, so here I am on the website.

  • I'm gonna go down to you.

  • Net.

  • Looks like there's a placeholder image right now.

  • Um, let's quickly go to Acknowledgments, huh?

  • People think this is, like, so in progress.

  • Look, can I get to the old website just for the purpose of this video?

  • Let's just see for a second, I think What if I change this tomo five Js dot or ge force it to do that?

  • No, no, no, that's prg.

  • Yeah, all right.

  • I'm just gonna I'm gonna, uh, live with this right now.

  • And the construction noise always the construction noise.

  • How'm I doing Timewise Okay, Not so bad.

  • Could be worse.

  • Okay.

  • Thank you.

  • To Russ.

  • Super Tara as well, Huh?

  • All right, usage.

  • So I'm not gonna worry about going into the reference in detail.

  • I'm just going to we'll cut to this when we make the tutorial.

  • Uh um please do a video about I don't know what that is.

  • I'm gonna have to look up what?

  • That is.

  • Um all right.

  • Um, let me just let make a pitch to support open source software development.

  • Lots of wonderful projects out there that could use your support as well as me.

  • Thank you.

  • Um, first thing I want to do is navigate over to the Ml five Jess Web page, the reference page for the U Net model.

  • Now you know I should be clear.

  • Her unit is you.

  • Net is a machine learning architecture.

  • There's no what I should be.

  • First thing I want to do is eyes navigate over to the Ml five jazz website and look for the U net reference page.

  • So on the reference page.

  • Also find some information about the about the background behind this functionality in Ml five as well, a sort of like a quick start guide for what the basics of code works, how the code works and more stuff.

  • So what when they should be clear?

  • Here is that you net is The machine is the is the model architecture.

  • So on its own, it isn't anything but Sayyed trained a model based on faces.

  • So in ml five, to access that pre train model is the only one right now in the future, there might be more that you could put in here.

  • You'll say ml five dot you net face, so let's actually start putting that into our code So this is what I have so far.

  • I have a P five sketch that loads the video from the Web cam.

  • You can see the green screen here, and it's displaying it here and then I'm redrawing into the canvas.

  • So what else do you want to do?

  • Is draw the segmentation here and see and see the original video here, so I'll go back to the code.

  • A couple things to mention is going to make sure you have.

  • In addition to importing the P five library, you have the ML five library important as well, and now I can go over and start to edit sketch dot Js.

  • So back to the reference page.

  • I want to say this, and most likely let's just see if this works. 01:01:

Okay.

Subtitles and vocabulary

Click the word to look it up Click the word to find further inforamtion about it