 It now looks better that there is no background, so it looks more like an output Yeah, we don't like it's but hopefully they're not too intrusive to now delayed like by 10 seconds or so I probably could delay them further. I was just Playing around with the delay function, which is done on my server server and yeah This need to be made clear that that field order is only affecting the frequency domain We do it now. There's a description So they built for this lecture just from one to four just the sharpness of the band filter That's all the setting is at two and why don't I have a field of feeling? That's Now that looks correct This thing would not actually show you the value like the window size there that Little value there will actually change. This is in production by the way that for So you can go check it out In this you are real pop in the shit for you to play leave Can change your in different Types of weblets at the window size. Yeah, there's a lot to play with I might leave that There for now We have another blog hopefully all these blogs will eventually turn into web application So that will be more useful than just text. They will actually show Something useful like EG signal quality Yeah, we can do a lot we can do a web application there Yeah, this application is an early version So a simpler version of the EG data time and frequency domain is just scrolling through the file You can select the channel seconds same as the previous tool the newer tool and this one here obviously has also options to the Signal less noisy They play with that. Okay, the ads are really annoying should probably showing this on my Development server not in production. Hey, what else? Let's go to the bottom quickly jump around this one a Yeah, those five do not currently work So should I go out of these three any preferences for what I should be looking at first Let's open them quickly side by side Yeah, we had the original tool that used to work Hey, we have some description to it that we can pop into a Chage-upd might as well do it straight away. I'm pretty sure we had something already Eager extraction. So this is feature extraction from images Actually have some unique images that I could be Trying this on feature structure Right, we don't have a description That's that all the lab you the HTML. It's actually a whatever PHP. Yeah, I'm getting rid of them And let's have an info text First that we had the original video. It actually shows how It's supposed to work, but this text can be popped In here. We have the description Fast feature point detector detects corners in an image using the features from accelerated segment test FASD algorithm Fast identifies all interesting features in an image and selects features so that they're consistently detected Harris corner detector uses the Harris corner detector algorithm to detect corners The Harris corner detector is a mathematical operator that finds corners in an image Sheet to Marseille corner detector uses the sheet to Marseille corner detector algorithm to detect corners Okay, so I That's description enough. So we have our inputs and the type image options RS and sheet to Marseille. Yeah, we have that description in the bottom and We also need the raw data option This will be on off and the output or should we just display the raw the original image? No, let's just display the original image We'll have our outputs original Processed oh come on Get up Yes And process the edge right and that's some this Scripture So that now the question is a Should this be a flask application or not? Or can it run on the client side? Start the new Shit going to be for straight away. Yeah, let's see what it says Okay, many considerations ever science processing privacy risk built-in integration development consideration Yeah, on the client side there might be limited Capabilities, let's go with last yeah, and we already have like what the 8 9 10 of them really so it's just One should be easier. I'm also checking what the images we could use as examples basic structure So we have a web interface back-end the back-end does they get post requests Using open CV to implement the feature detection. That's great Saving the images temporarily. That's okay Maybe change it later. Let's play the output ending road What That actually is interesting and normally when you're talking about biomedical images, you're thinking, you know scans But yes, actually time series data can also be an image So it's actually could be quite an interesting thing to do because we could just go into one of the other a tools and Generate quite a few images and see how the image detector deals with them or should we go Imaging examples because the algorithm will be Steps to implement vision detection stalled flask opens a bit, but Headless I have flask install Yeah, I Ajax Did use Ajax in one of the application but not in a more recent one Shouldn't matter too much for e.g. this versus say a micro scans Just see what it says. Yeah, obviously you will process the image quite differently, and you'll probably look at them different features Perhaps a few is my studio data characteristics volumetric image processing a complexity of that Yes, that is more complex. I mean, um Um, I don't think I have 3d images. Anyway, that would be probably could try and get some 3d micro ct But at first stage it probably just gonna be to the images Yeah, and what's the final outcome? Is that just using it? For example the one the diagnostics what things are setting research development engineering context This will dictate the precision speed via one like the Many for educational purposes of mind. Uh, so speed is important And optimize for speed Now, yeah, where do we Get the images from Isn't important I have another all that called data A index HTML And let's do this structure of last application files and All this Don't need to spell correctly. Try to people understand Basically a main page Result display Okay, so not in a single page panel For a single page panel your application Like gs images, that's all the root definitions. It's a bit odd. Can I have us done that? That's the last application that we have made this one. Let's do it in the same way. IG spectrum and the noise Yeah, like this structure And this will be the images and we'll see where we get the images from in a second Right, so the app is called a feature extraction It will have an app by a requirements read me static Style is one. What's the standard name or main JavaScript templates index HTML. Now data This will be images the JavaScript It's not for e g and this will be image And let's do this Okay, I don't know why just rearrange the thing explaining each file. I didn't ask you to that's okay. Can we stop generating that? And I have those Those go next They generate all necessary code for the updated hd for well just html. I mean, this is the first time we're generating html. So Yeah, we're just generating the templates the JavaScript file. That's while it's generating Let's do the folders quickly or should we just copy them? And we have the info file Let's copy those quickly my app requirements We will need to regenerate those Let's just create the m using the DOM content loaded That looks good I file and straight away correct this Into the new application name generated some Python code Just the template Right, we're really generating the html. Yes, we want to separate css And probably after css we'll have to update the html again Obviously We're not ready to deploy with we just started literally With this application Okay, where is this? Yes, all the body h1, h2 h1, h2 should be commented out Right html template Yeah, just checking that it doesn't have any styling or JavaScript in it Because we would like all those be separate Okay, I don't like this already because I said the style css Is not in css folder Content ID Form Input file image And we're using Plotly JavaScript main A JavaScript can be renamed main JavaScript Need some images and data at least one if we are doing EEG Segment we could actually This is actual an actual seizure How about download plot as png Let's call it seizure One png pop it into data info Get more prompts Make simple image To use And then Um, we haven't done Python yet So we're using flask CQ file and what just go Waited for now Okay, example the images in This file Okay in html Where do we edit? It looks like again Yeah, choose file Definitely Okay All the image files are in this folder And choose First file by default So this actually does not Meta Just pop another Image in there We actually had this Images for The wavelet image compression Which one is good? Yeah, the first one Compress wavelet Right, so this should be In static Images First one MRI brain SA02 Yes, this one Static images Okay, let's rename this into Images And put it in this static folder Name this one A01 seizure Yeah, so it's just on top of the file A copy path It's meant to be relative path Static images seizure one or one seizure is static images folder Choose first file by default Had a drop-down menu to select the image file Yes for that The vector type we wanted to do that Yeah, that's GitHub co-pilot And those are good Why are we using web browsing? Should be just using the standard GPT-4 default Filter out files that are not images if necessary Yeah, we don't really have to do it We'll note at this stage To change our HTML Render template Yeah, we don't need that secure file in what Cause the images will be just on the server Get rid of that image folder That's right Yeah, we don't want a dummy extraction function We want a real extraction function Yeah, let's update this Witen code HTML Next Doesn't mean we don't need plotly Deep plotly chart Right, so we have the file Choose an image Choose the vector And why is it not populated? Why is it something that is flicking? Question is that if it's rewriting the whole A flash cap would just be it The debugging is true Yeah, we're still debugging Hey, if you have snippet this here Let's create an index HTML template for the drop-down So HTML And bum-bum-bum-bum-bum Scripts Drop-down menu That's the same image Yep Yeah Option It's just the same I don't get it Why is this not Opulated Images study images And the index HTML I'm going in circles here Like this is just the same Don't get it All right, I know what the problem is Because I'm not running it as a flask application I have to run it as flask That helps, doesn't it? File not file Static images Static images Static images How come the file is not found? Copy search And the folder Copy relative file Example Yet Okay, quickly fix this for me So we can actually move on to the actual development Is the whole thing? I used to just leave me to the workplace Right, okay, now it's working Okay, we are in business Sure, images are displayed I'm using co-pilot all the time It's just, I actually had to change its settings as well Co-pilot No, I'm using it all the time I didn't realize it had quite a few settings that I Yeah, I think it was actually set up for developers Where you're doing everything myself I'm not a developer So now it's actually should be better I haven't checked it yet Make sure images displayed in the HTML files So I can try them both side by side I would actually love to ditch chpd Well, because it's more expensive I don't really need it If I knew how to use github co-pilot properly So for example, yeah, I'll try them both To say, obviously, it's directing us to change the HTML Because we meant to have these two images original and processed So we have the drop down menu Drop down menu Right here A place original and processed images side by side Yeah, that was Yeah, now it's thinking and it generated the whole code Give me two options Yeah, I like the second one Accept for tab Save Didn't I say side by side? Definitely not side by side And original image And well, can't it see script HTML Make sure the images are displayed in the HTML file Yeah, so this is a much better example Place it together Right So yeah, you can tell GPT-3 is actually having context for the whole thing I'm surprised Github co-pilot has access to all my files and everything It's in literally in the development suite But when it's giving me advice, it's giving me generic stuff Whereas GPT gave me the actual file with the right folder And it actually works So, yeah Being with 3.5 is, yeah I think it's a problem that people use it And then they get disappointed from it The main reason I'm able to use And I don't know why it's in web browsing Is the fact I can start fresh for now Do it quickly because I'm running out of time But if I'm in default GPT-4 The main reason I'm able to use it is because it can store The whole context for all the files So if I have all this one, two, three, four, five, six files If I pop each one of them I'll go in for last The order is important Index, HTML first So if I start from scratch The good thing about it Is it the right file? Yeah Yeah, and I'm just telling it to read the code And don't generate anything else And that's because it then stores everything in In its memory somehow You can use these lines to work They quite well And they pop this again So it's actually, yeah, it says, yeah, great The code's provided So it has it in memory It read the Flask application And that's great Stiles says I can sort it out later as well A JavaScript is most important We use the same line Confirm you read it And do not generate anything Any code or anything else Copy all Yeah, so GPT-4 is able to To keep the context of all the different files So we're talking Well, we're talking 100 lines in HTML 30 lines in Flask so far 40 in JavaScript And now I can actually ask it questions So I'll go into my application Yes, that HTML that GitHub Copilot generated Didn't actually work Now I'm actually trying to find it and get rid of it And this one here It gave me very generic stuff And yeah, I just don't know how to use it The whole point for me Is that I'm able to Where was I? Yeah, read the JavaScript It gave me some error or something So generating another prompt You don't have to worry about spelling Just pop it in as is And it's giving me Yeah, it has the context So that's the stuff that I already have And then it's telling me what I need to change And it looks correct I don't have the processed image yet Hopefully we'll have it soon enough Save Yeah, I don't have the processed image That's a problem And it can do processing for me Because it had the Python code as well Okay, I'll continue this later See you in a bit