VGG16 Neural Network Visualization

[@fayezsalka]

This has to be one of the most inspiring ML videos on YT

[@extropiantranshuman]

I love this utter visualization of what the neural network's going 2:00

[@cipherxen2]

Thank you, thank you, thank you very much. I know how everything works, but seeing that in action is another thing. Thank you once again.

[@paritoshverma8655]

Videos take lots of efforts, to give a very practical understanding of things

[@shivampatel8928]

This is amazing... I will pursuade many people to come here and and watch this

[@shonash2651]

Глаза радуются, великолепно проделанная работа. Горжусь соотечественниками !

[@complexobjects]

Unbelievable visualization. Conveys the beauty of CNNs while showing an key aspect of how it works. Thanks.

[@berkay2608]

swear to god this is the best production in a ML tutorial. too much inspiration, thank you so much.

[@berkay2608]

i was like blind the whole time that i was working with cnns

[@Darthvanger]

The birth of Skynet... :) Beyond amazing visualization! Even better than OpenAI microscope :)

[@gpt-jcommentbot4759]

The visualization was mindblowing

[@Buddhxst]

✨どうもありがとうございました ✨

[@SploxLabs]

If i had to ask of anything of the great and powerful algorithm, please recommend this video to as many users as possible!

[@venture_lab]

This video needs more views! Excellent work and quality presentation!

[@arabnaouel289]

the best visualization that i never see thanks very much i'm in love with this demonstration

[@ketongwang3177]

This is definitely one of the best I've seen! Liked immediately. But maybe using more labeled fruit images from Google will better demonstrate the power of VGG16? Since one single fruit is used for each kind, the shape might be easily learned and recognized by the the networks. A lot of prep effort tho! Again, very nice work and can't wait to see more. Cheers!

[@pedrovergaraurquijo8551]

Completely agree, I would like to see how his model would generalize.

[@lavinnasays]

Dude. This just blew my mind. Respect!

[@sumod12]

So sad there is no love button ❣️

[@tellmebaby183]

beautiful , I have one question - what weed or shroom you do!

[@simplexination9837]

Like a child is born and learn first things. Don't know why I'm cring may be due to happiness.

[@hemaswaroop7970]

Wowww!!! It's a sheer visual pleasure.😍😍😍👍👍

[@notskinner2304]

Everybody teaches you what to do but the how part you have to figure yourself out.

[@AECTechJourneys]

Incredible work.

[@thangnguyen8063]

this is the best thing i've ever seen

[@8eck]

Whoa! How much convolution layers. That's some real pro level. :D

[@minjoonkim4766]

It's really great. I watched the best video of my DL life.

[@khaya-ludidi]

Really awesome, but the video is rather hard to watch if the majority of it is whizzing past all the layers at nauseating speeds...

[@samunel33]

This is crazy impressive 😲

[@allessandroable]

What are those, activation maps? Wouldn’t b e more useful see the learned filters?

[@adityagautam4974]

This wouldn't solve the real world problem in efficient way since in real world you would have different background, different size, color, shapes, and half/partial image, multiple fruits. But this is very good for basic understanding.

[@EmptyNonsens]

damn, it must have taken months to prepare

[@artinbogdanov7229]

Great video!

[@datahouse24]

Great content 😯😯❤❤❤❤

[@SploxLabs]

HOLY HELL THIS IS AMAZING!!!

[@zaqueudovale852]

Great, amazing, awesome!!!!!!

[@mymycat187]

why this incredible like this has so little view, please dont make him give up to do this kind of video , super great video

[@daninjeon5566]

고마워요 행님.!

[@user-or7ji5hv8y]

But can your model generalize well given the context for which data was made and collected?

[@baptvitai]

Contras !!! Hell of working !!!... I'm doing a similar project with XAI tecniques for Skin Lesion classification.

[@lesliechiang1710]

Very nice work!

[@InNOutTube]

I was about to sleep, but now I'll have to develop VGGNET16 model then I'll go to bed :)

[@rodrigofarias900]

hi Denis! Great video. Can you give some reference (link) that presents a simple example of CNN, from scracth. I'm trying to learn, but the examples i found are not good at all.

[@antond1021]

Did no one else see the typo at 1:52? The NN has a 'Pooing' layer lmfao

[@anguswu2685]

Anton D I guess it makes sense too. Less stuff comes out than what goes in when your body is absorbing nutrients. Same with NN 😂

[@konataizumi5829]

"The model will be trained on GTX1070 until 0.99% validation set accuracy" bruh lol, that seems kind of small?

[@asmarly]

i think they mean 99%

[@wishIKnewHowToLove]

Молодец, Денис!)

[@DheemanSaha]

Just keen to know how did you create the visualization among the layers.

[@alim.hosseini4638]

Good Job

[@shirleysun3539]

inspiring. I just wondering the video of fruits was made by yourself for obtaining more data or, offered by kaggle and open to every team?

[@DenisDmitrievDeepRobotics]

It was done by me from scratch.

[@DANstudiosable]

How did u show the layer to layer insights of the input image?! Please explain. Or try to make a tutorial on it, it'll be really helpful!!!!

[@macsenwyn7223]

Do this with no music and show every layer of each fruit please

[@NovaWarrior77]

AWESOME!

[@jackzeng7096]

So many layers! maybe it is a bit overkill?

[@YeekyYeeky]

superb

[@AB-dw8vo]

how was the visualization of layers done?

[@biiingos47]

At each layer in the forward pass save the activation directly and compile it into a video I guess

[@thoughte2432]

60fps at 3:00 would have been suitable, nice work nonetheless.

[@hyungsukkim3396]

cool video

[@ostadnavid]

What tool did you use for animating conv layers ?

[@8eck]

Is that a thermal imager in conv levels? O_o

[@farshadnematpour4016]

Can you tell me how to visualize the model architecture with this method !? how can I do that !?

[@DenisDmitrievDeepRobotics]

Keras, Python, MaxScript, 3DSMAX

[@Diamond_Hanz]

beautiful

[@kolokolo2365]

Great video. How did you visualize the feature maps ?

[@JAYMOAP]

Well done

[@nguyenichthanhtu8502]

amazing video

[@picklestirfry]

Impressive, but can it tell me if something is a hot dog or a-not hot dog?

[@sahilaggarwal4973]

Amazing, man how you make these animations, please let me know please. Amazing work

[@shis10]

Awesome

[@jayshaligram4474]

hello Denis! amazing video! what is the maximum catagories that can be used as input? right now you are using 10 different catagories. but can this be done for more catagories? like 100, 500, 1000?

[@DenisDmitrievDeepRobotics]

Sure

[@torcher5023]

How can 1 channel filters be visualized as rgb images?

[@omidalekasir4736]

I have a question! Can we run an industrial process with a program written with python? Is it safe and "professional"? (ppl say use LABVIEW)

[@DenisDmitrievDeepRobotics]

Yes. As long as we ensure stability and other related things:-)

[@ابراهيمالعلواني-ع5ض]

can i get data set please??

[@johanneszwilling]

...😀 You are back! Any news on a tutorial? Or if possible: Sharing the code with documentation? Would a donation help this along? 😎 Regardless: Thank you for making this!

[@arifulislamleeton]

Hi I'm Ariful Islam leeton im software engineer and Students data since And technology

[@mayankmishra3875]

0.99% accuracy?

[@eranfeit]

Hi , How did you visualize the layers ?

[@dariomendoza6079]

OpenGL for visualization?

[@DenisDmitrievDeepRobotics]

OpenCV, Python, Maxscript, 3dsmax

[@thehonestdude1067]

that's really cool 😃😃😃

[@sanskarshrivastava5193]

God damn ! ❤️❤️😍

[@muhammadadhityareskipratam4502]

can i get the code?

[@nissimtrabelsy3957]

wow

[@humani262]

youtube's moron ai, where are the views?!

[@AirMarty78]

-1 for music...

[@addyad_yt]

beautiful

[@omidalekasir4736]

Awesome!