The U-Net (actually) explained in 10 minutes

Рет қаралды 132,120

Күн бұрын

Пікірлер: 129

@salmanzafarsatti1346 Жыл бұрын

man, this video is such a great explainer. I was confused about the use of skip connections since a long a time, but he explained the intuition behind it very nicely.

@Anton_Sh. Жыл бұрын

This architecture is one of the truly brilliant ones in the world of deep learning in terms of its simplicity and efficiency.

@mayankukani9600 Жыл бұрын

Why didn't I find your channel before. Please upload more content, the best content on Deep Learning I have seen.

@rupert_ai Жыл бұрын

Thanks a lot :)

@nikhilchouhan1802 8 ай бұрын

You might not find my comment since the video is too old, but man I just want to thank you for this video. I am a student who has always been interested in computer graphics and related fields like game engines, physical rendering, ray tracing, etc, and jst didnt get the ML/AI hype everyone was on the past 2 years. I only ever managed to study ML basics for 2 weeks before I left it for good. But recently I got in a team where my friends were working on CNN based projects, and that made me learn about many basics about NNs and DL. This explaination for Unet seals the deal for me, and I will strive to work on integrating my two interests into one and hopefully create something I love.

@jayhu2296 9 ай бұрын

your explained under 10 minutes videos are goated

@mridulsehgal7773 7 ай бұрын

The best ever video you can get on Unet explaination

@Atreyuwu 4 ай бұрын

Not even close lol

@thebakareview8009 9 ай бұрын

This channel deserves more subss!! Great content and delivery :)

@puekai 9 ай бұрын

Still don't know how it works

@vardhan254 6 ай бұрын

me when reading goodfellow all night

@prathamghavri8591 5 ай бұрын

@@vardhan254 dude hows the book ,what would u suggest so that one has a good read

@arf9759 5 ай бұрын

No one really knows how/why a CNN works!

@cupckae1 3 ай бұрын

@@arf9759 Exactly and these make sense to machines during training(When backpropagating errors). this is the reason filters are initialized randomly and are trained.

@haruthunanyan7387 3 ай бұрын

@@arf9759 I don't know what are you referring to, but there's actually a mathematical proof why conv NN's are used in image classification. Check out geometric deep learning by michael bronstein

@JohnZakaria 10 ай бұрын

This was the best unet explanation I have ever seen

@pushkar9021 Жыл бұрын

Continue this series, very helpful

@Natstranaut Жыл бұрын

Oh my god man. Awesome videos. Keep it up, I'm really enjoying them!

@ABCEE1000 3 ай бұрын

Man i like you ! . you are the best ! how you simplify thing and how you are careful to deliver the idea perfectly >> please keep this great presentation up >>

@LucaBovelli 7 ай бұрын

dude thankssssss i thought this was another one of these things thatll take me 2 hours of youtube to *not* understand, but u saved me

@NoOne-p3e Жыл бұрын

Extremely useful for beginners like me. This is very good

@jacobidoko3924 11 ай бұрын

Yooo...this is quality content right here. Thank you so much for putting this out

@shubhamarle96 8 ай бұрын

thanks for the video, I am trying to use U-net for anomaly detection in time series and your video gave me the idea.

@miguelxplayer9641 9 ай бұрын

Dude, you're great. I'm from Portuga 🇵🇹 🟩🟨🟥🟥and I'm learning Machine Learning and Neural Networks. Thank you very much! I loved how you teach. You are intuitive and dynamic. A person is learning a difficult subject and still manages to laugh when watching the videos. I loved. I already subscribed and liked. I'm going to watch more of your videos now. Hugs from Portugal😉

@jsparger Жыл бұрын

This was extremely helpful. Thank you

@caleharrison5387 3 ай бұрын

Thanks, this is really good. One thing that would be helpful is if the example was itself convoluted like the algorithm, to make easier to visualise the algo.

@DannyGeisz-vb2dt Ай бұрын

Thank you Rupert! Excellent, excellent explanation and intuition for this :)

@rippingmyheartwassoeasy 10 ай бұрын

Thank you for creating this video! Its the best explaination of how a U-Net works that was easy to understand. The visual animation is superbly done!!

@niralpatel5889 8 ай бұрын

This was great, would love a video on diffusion transformers! It looks like they are taking off and replacing U-Net's as the backbone to new diffusion models.

@hemalathat8773 4 ай бұрын

I LIKEED THE ANIMATIONS AND YOUR PTESENTING STYLE IN THE VIDEO. THANKS.

@yyww4267 Жыл бұрын

Really impressive vedio! And fun work at the end!!!!! LOVE LOVE LOVE!!!

@rupert_ai Жыл бұрын

Thank you very much! :)

@rogercomix5648 2 ай бұрын

I liked it but you did not explain the role of the 3*3 kernel, and how it scans the pixels of the image at each layer, and the reason for the downsampling is because it is more expensive to increase the size of the kernel at each layer so we downsample the image so we get the same relative size differential as if we did increase the size of the kernel. Apart from that, it’s brilliant.

@vgtgoat Ай бұрын

I'm not an expert but here's what I understand. The conv filters on the earlier full resolution image will learn highly detailed features such as edges. The conv layers run on downsampled (lower resolution) images can't see edges because they're all fuzzy now, so they will learn more large-scale features, such as shapes, then objects. As for how the 3*3 kernel (filter) scans, I believe it's just a standard convolution which you can learn from other videos.

@s4lome792 7 ай бұрын

Clearly explained. What caused my consfusion in the first place is, in the graphic in the original paper, why does the segmentation mask not have the same dimensionality than the input image?

@nathnaeldereje5127 16 күн бұрын

Great video, I like the demonstration. You just earned a subscriber 😆

@dhanushs4833 4 ай бұрын

great vide mate , would love to see more brilliant stuff like this❤❤

@oblivitus. Ай бұрын

brilliant! thank you for this illustration!

@TheHopeOfTruth 9 ай бұрын

Thank you for great explanation.On basic level it helps better understand unet

@liliznotatnikiem6755 4 ай бұрын

I’m interested at multiclass problems (recognising bike, human AND house). Also what would you choose instead of confusion matrix?

@ubanaga 11 ай бұрын

Very nice my friend, this has been most helpful

@nguyenangkhanh4971 Ай бұрын

great, hope you continue with the videos

@pranavgandhiprojects 5 ай бұрын

Hey just show this first video from your channel and immediately subscribed to your:) Great explaination with visuals

@faaz12356 5 ай бұрын

Very useful and great explanation.

@coffeestudi0s Жыл бұрын

Yooo the effort haha. Amazing Video!!!

@terjeoseberg990 6 ай бұрын

You didn't explain how the skip connections are connected across. What is the data that's transferred and how is it incorporated into the output half of the U-Net?

@sakethsreeram6981 9 ай бұрын

Great presentation!, Easy to understand

@pratyushsahoo4948 9 ай бұрын

Absolutely amazing work 🎉

@gokulsaisrinivas5312 Жыл бұрын

very good explanation of U-NET

@xarisalkiviadis2162 9 ай бұрын

Amazing video, cleared everything!

@hexeldev Жыл бұрын

This video has been extremely useful. I subbed.

@transcendingvictor 10 ай бұрын

Thank you very much for the time put on doing thisvideo. Interesting and helpful :)

@VikashSingh-vd9cp 6 ай бұрын

bestvideo for understanding U-net model

@DanielaFrankl-l8t Жыл бұрын

Woooooow! Finally I understood it , really great explanation, thank you

@nagham96 Жыл бұрын

Thank you that was so helpful and cute! 🤩

@willlowtree Жыл бұрын

i love your presentation style

@ABCEE1000 3 ай бұрын

whould you please make a presentation on 3D Unet . that would be really appreciated

@dfparker2002 Жыл бұрын

This explains inference (I think) by decomposition (dividing) and recomposition (adding) images. Is that accurate?

@Ngochi-ff7hk 8 ай бұрын

I still don't understand that the output is x2 or x3 or x4.I don't understand why that is the case?

@atifadib 7 ай бұрын

If you want to just use the Decoder how would you do it?

@aligreen786 11 ай бұрын

Very nice explanation. Thanks a lot.

@BooleanDisorder 10 ай бұрын

What's the background music called in this video?

@citiesoftheworld381 29 күн бұрын

Shouldn't concatenation lead to an increase in channel numbers? I don't see that here.

@JohnVinchi-bk2dw Жыл бұрын

this is extreeeemely helpful,and funny

@rupert_ai Жыл бұрын

Thanks John!

@_the_one_who_asked_ Жыл бұрын

Hi, thank u for this video. can u pls do a video to explain YOLO?

@tho_norlha 27 күн бұрын

thank you very much for your explanations !

@ny8828 Жыл бұрын

hi its very helpful, how can I reach the PowerPoint of it?

@IzludeTingel 29 күн бұрын

any AI out there able to use an "exact image" in a newly generated form? for example, i draw something, i want AI to make that exact thing (perhaps a cartoon animal i created) and make that exact character doing something cool so that i can make it into a shirt. i have characters, but lack the creativity to make attractive shirts out of them. i dont want it to be a "similar" looking character that's completely different from what i created, i want it to be my exact characters.

@amolkumar1538 Жыл бұрын

This is Just awesome, great video

@Nerthexx Жыл бұрын

If downsampling works by max-pooling, how does upsampling work? In traditional image processing, we would just interpolate image colors, but how does the network apply it's "convolution" in this process? I would understand "deconvolution", but in my mind it wouldn't work here.

@AyushGupta-fv1lx 8 ай бұрын

May be Transpose Convolution

@notrito 7 ай бұрын

If anyone wonders how to concatenate the features if they don't match the size... they crop it.

@SyncMicroaga 10 ай бұрын

Hi. I find the video very interresting. As I'm at the begining, i'm little confused. please, can you also propose a pdf file ? thank yu. Nicely

@Topninja6 8 ай бұрын

Thank you so much. Now I just need to figure out how to implement this for my project lol

@mincasurong 7 ай бұрын

Great summary, Great thanks

@HelloIamLauraa 5 ай бұрын

best explainer!! great video, I had an "aaaaááaaa" moment at 8:05

@alirezasaberi6383 Жыл бұрын

awesome! can you calso make similar (actually) for Unet++ and Unet3+ please??? thank you so much.

@rupert_ai Жыл бұрын

Glad you liked it! Its not currently on my list of to-do videos as I like to cover the most popular fundamentals at the moment, but I'll let you know if I get around to it! :)

@ShwethaVme23b073 8 ай бұрын

wow awesome video and explanation

@ozzafar1982 7 ай бұрын

great explanation thanks!

@siddhantpassi8237 2 ай бұрын

Amazing video!

@MacProUser99876 10 ай бұрын

nice explanation. but why distracting background music?

@endlesshybrids 7 ай бұрын

Agreed. Good explanation but I wish people would stop using background music.

@r.walid2323 8 ай бұрын

thanks, good explanation

@TechHuntBD 7 ай бұрын

Nice explanation

@shinobidattebayo7650 4 ай бұрын

nice effort, but the sound of music is distracting.

@vijaykumarb9622 11 ай бұрын

Great Explanation.

@kiraqueenyt5161 Жыл бұрын

such a well made video

@1.4142 Жыл бұрын

Dalle 3 is coming to gpt 4 and it can write text!

@Lautaro04000 8 ай бұрын

nice video, very helpful

@MrMadmaggot Жыл бұрын

Now how they coded it?

@rupert_ai Жыл бұрын

Hahaha well there are actually plenty of online code implementations available but I will see if I can get round to a code tutorial on the u-net sooner rather than later!

@rishabhbhardwajiitb178 11 ай бұрын

@@rupert_ai can u provide one

@ajipboy 9 ай бұрын

bro , immediate subscribe!

@gregorioosorio16687 Жыл бұрын

Thanks for sharing!

@jaybrodnax 7 ай бұрын

I feel like this is more a description to experts than an actual explanation of how and why it works. Questions I'm left with: What is the purpose of downsampling/upsampling (I'm guessing performance?) How is segmentation actually done by the u-net? How is feature extraction actually done? What are max pooling layers? What does "channel doubling" mean, and what does it achieve? How does the encoder know "these are the pixels where the bike is"? Why is it beneficial to connect the encoder features to the decoder features at each step, versus in the last step? How does unet achieve anything other than downscaling/upscaling performance efficiency? Where are the actual operations to derive features? How is u-net specifically applied for various use cases like diffusion? What does diffusion add or change, for example.

@abansalah4677 7 ай бұрын

(Disclaimer: I am a beginner, and this is not intended to be a complete answer.) You should read about convolutional layers and pooling layers to better understand this video. At any rate: A colored image has three channels: R, G, and B. A convolutional layer is specified by some spatial parameters (stride, kernel size, padding) and how many filters are there - the number of filters is the number of channels of the output. You can think of each filter as trying to capture different information. Doubling the channels, therefore, means using double the number of filters when using a stride of 2. The segmentation is done just like any ML task - the training data consists of pairs of images and their annotated versions. I think it's often hard to decipher the inner workings of a particular neural networks, and your question can/should be asked in a more general way - how do neural networks learn?