Neural Networks Part 8: Image Classification with Convolutional Neural Networks (CNNs)

THIS IS AMAZING, Josh. Thank you so much. now I understand the concept.

What I could make of this video is that once the filter, its bias and are tuned through back-propagation and appropriate activation function is selected, after max pooling its vector is fed into the ordinary neural network which learns to classify whether such kind of matrix could come from X or O.Have i understood correctly..?
Also the max pooled vector/matrix will fill 0s and 1s in relatively unique ways (for X and O) depending upon how the filter has matched to the original image.That forms basis for classification..!!

you are goated no cap

❤ nice video 💯💯👍👍👍🫡🫡

Wow!
What an explanation!
Thank you so much for all these wonderful contents.
Keep it up! 🎉

Hey, Thanks for the simplified explanation.... it's too good!!

Thanks!

fantastic explanation!

I can't imagine how you can explain so simply ...hats off to your work ..great and superb explanation...need lot of statistical videos like this

Thanks!

i did'nt understood(after multipliction of weight), how it is working how we are getting 1 or 0?

loved it❤❤❤❤

Hi Josh! Based on the video, is it right to say that the number of filters used is only one? I'm really glad I found you. Your work has been tremendously helpful in understanding these topics!

I have a question, I tried using what you did in this video for grayscaled images, but it didnt work very well(to be specific,the first filtering). What do I need to change for it to work with grayscaled images?

Why did we use only 1 feature? It isnt always the case right?
Also both images had the feature(kernel) 2 times, how could it differentiate betwenn them??

Excellent video, To the point and complete description. I was actually looking for a good course on NN. I will use the playlist provided in the channel

Thumbs up even before starting the video cause I'm sure it's going to be awesome!

I was watching these years ago and still watch them now when I want to review. One question I had on this video:
When making a filter, you showed us adding a bias term to the output. Since the bias term is in green, it sounds like it's something that should be computed via back-propagation like the other weights and biases. But I don't understand how you can do that given the input nodes to the neural network don't start until after the filter and max pooling step

Thank you sirrrrrrrr

the best channel ever.

Hello, thank you very much for this video! How does this procedure work for RGB images, i.e. images with 3 layers instead of 1?

Thanks dear Josh for the quality content as always, I have a question though about activation functions in the output layer, is it common to train a neural network without an activation function in the output layer? For example in your video we could maybe backpropagate -0.23 and 0.2 through the network to get 1 and 0 for the predictions, without ever running them through a softmax/argmax.
I am asking because I saw some code somewher doing it and I oculd wrap my head aroud the math, what do you think?

I am an MSc student of Data Science. I learn alot here than in Class😅

Really helpful! Just a shame that it doesn't have more view :o Subscribed.

Question for anyone watching now: It was stated that the filters pre-training and after-training are not the same, yet I'm not sure how do the filters themselves get optimized? Because the input nodes come after the filters, activations, and pooling, and the only thing optimized after the input nodes are present are the weights and biases that lead it to some output. So how do the filters go to the diagonal bottom left to top right from some other initialized arrangement of the filter? Any help is appreciated.

sir, recently i have bought an illustrated guide for machine learning but in this book, there are lots of topics uncovered like k nearest neighbor, PCA,etc

Why CNN doesn't show up in your The StatQuest Illustrated Guide to Machine Learning (PDF)?

who is J. Butt in the credit at the end

Thank you so much. I could not gave asked for a better explanation?
I have a few quick questions though:
1. How do you decide a filter?
2. What is the impact of colors on filter?
3. How could this theoretical learning could be turned into hands-on experience to connect our mind neural nets? :)

how can we estimate weights and bias?

💚

Awesome explanation. Thank you so much.

Small bam 😂

thank you for such an interresting video

can anyone explain the dot product used in the video? it looks different than matrix multiplication.

Man, you're the best!!

This is the best neural network layers explained in the entire video community

This was an amazing video! However, I'm a bit confused about how higher complexity neural networks work? From other sources I have read that CNN's can have multiple convolutional layers, but this doesn't make sense to me. If convolution is used to get a general map of features from the input, and you use max pooling to get values which you input into nodes and you calculate weighted sums etc how can you reapply convolution in a further layer?

Abicim yemin ederim hayatımda bu kadar iyi anlatım gördüğümü sanmıyorum

You're literally a saint for this content! Thank you very much.

Super cool as usual. But it would also be super nice if you explain the case where there are multiple filters. The topic CNN deserves more quests:)

Thanks for your amazing videos! Could you possibly make a video explaining Graph Neural Networks?

BAM!!... You explained very easily and clearly.... Bam!!!

The best CNN explanation I've ever seen. However, i have one question about the part of classification of 0 or 1. As a classification problem, why there is no sigmoid or softmax function used in the last layer, are we just using the raw output to make prediction?

"I don't know how much time does Artificial Neural Networks take to train, learn the input data, But you are putting more efforts and it taking much time in your training time".Thanks to your efforts sir.

, your videos really explains very well and it helps us in visualizing easily.

wow, you're the best at explaining's things for easy understanding. "Simply Great"