2D Convolution Neural Network Animation

#2DCNN #ConvolutionNeuralNetwork #Deep Learning
This video shows the complete architecture of 2d cnn through animation

[@anujshah645]

At time stamps 1:23 to 1:28 the output of the second convolution would be 61x61x3 instead of 61x61x4 and the output of the second maxpool would be 30x30x3 instead of 30x30x4. Thank @sunnivasteiro532 for identifying that

[@ParthivShah]

nicee work.

[@TestMyHomeChannel]

I loved the animation. Would be quite helpful for teaching. I wish I had seen it 2 years ago when you posted it!

[@satou-mp6fr]

What filters did you use in the two Convolutional sessions ?

[@kuldeeppawar4028]

superb video, I just want to ask one question, initially when models starts training on dataset which values are present inside kernels? for ex. for the first conv layer we are just writing the kernel size as 96 filters of matrix 11*11 but in actual using which values we are comparing image pixels with kernel ?

[@SreyaB-rl8qd]

Hi, The animation is so good we can understand what is happening in the cnn network currently there is one more new network called Involution neural network which is location specific and channel agnostic where as cnn is channel specific and spatial agnostic so can you do a animation on inn also please..

[@alvinrachmat9328]

Great animation

[@anujshah645]

Thanks for the feedback

[@bdagro6782]

Best

[@Thanhnhon171]

Thanks for a great video. May I ask you a question? Four small matrices at the left corner in the convolution operation slide are Kernel filters, right? So are they different? If yes, what's that difference?

[@klaotische5701]

The filters‘ weights will be determined during training, it won’t be known in advance.

[@sunnivasteiro532]

Nice video, should the second convolution layer be 61x61x3 rather than 61x61x4? Since you applied 3 filters? I am new to CNNs

[@anujshah645]

yes you are right. Thanks for pointing out. Till 1:23 the values are correct, after 1:23 there is a typo. The output after second conv should be 61x61x3 and henceforth the output of second maxpool should be 30x30x3.

[@caiyu538]

This is a great animation to explain CNN neural network. Could I cite and show your KZbin video in my presentation in a medical related conference meeting? I will show the audience the KZbin link.

[@anujshah645]

yes definitely you can, you can use the content as you like as long as you cite it. Also FYI, these are preliminary versions of CNN. there are multiple changes in modern CNN, also highlight those to your audience. Also, this animation is a summarization of step by step tutorial I made 6 years ago for the earlier version of CNN, if interested you can have a look - kzbin.info/www/bejne/q2mliJKcoZuKiMU

[@caiyu538]

@@anujshah645 thank you so much. My audience is not familiar with AI. Your animation is a great tool to explain to them.

[@navidaleali5435]

❤👌

[@tintintintin576]

Superb video, gave me a lil more clarity... Can you please tell which animation tool you use?

[@anujshah645]

Thanks. I just usee microsoft powerpoint animation feature and for screen recording again I used powerpoint inbuilt screen recording feature

[@tintintintin576]

@@anujshah645 thank you so much.. Now I got it, I guess.. In case I'll find some difficulty I'll come back to get quick input from you. Big thanks 😇🙏

[@anujshah645]

@@tintintintin576 no worries, I have realised over the course of time that we can do everything with powerpoint, but it needs lot of patience and time

[@tintintintin576]

@@anujshah645 Interesting! Thank you so much for sharing this. I'm looking forward to trying out a few things...