What is CNN in deep learning? Convolutional Neural Network Explained
Vložit
- čas přidán 27. 07. 2024
- What is CNN in Deep Learning?
In this video, we understand what is CNN in Deep Learning and why do we need it.
CNN (or Convolutional Neural Network) is the building block of all Computer Vision applications. Applications like self-driving cars, object recognition, face recognition, etc.
There is a limitation of simple Neural Networks when it comes to dealing with images. They become extremely slow at training and processing images. And the number of parameters to train will also be very large.
So, to overcome this limitation, we use Convolutional Neural Network In deep learning.
Watch the video till the end to understand what is CNN in deep learning.
➖➖➖➖➖➖➖➖➖➖➖➖➖➖➖
Timestamp:
0:00 Intro
1:27 Drawback of ANN
3:12 Convolutional Neural Network
5:34 End
➖➖➖➖➖➖➖➖➖➖➖➖➖➖➖
Follow my entire playlist on Convolutional Neural Network (CNN), because I provide a very detailed mathematical explanation about every model, along with practical implementation.
📕 CNN Playlist: • What is CNN in deep le...
At the end of some videos, you will also find quizzes that can help you to understand the concept and retain your learning.
➖➖➖➖➖➖➖➖➖➖➖➖➖➖➖
✔ Complete Neural Network Playlist: • How Neural Networks wo...
➖➖➖➖➖➖➖➖➖➖➖➖➖➖➖
✔ Complete Logistic Regression Playlist: • Logistic Regression Ma...
✔ Complete Linear Regression Playlist: • What is Linear Regress...
➖➖➖➖➖➖➖➖➖➖➖➖➖➖➖
If you want to ride on the Lane of Machine Learning, then Subscribe ▶ to my channel here: czcams.com/channels/JFA.html...
Thank you very much bro, amazing explanation
Awesome content.. Loving your videos..
This is a beautiful picture of mine ! got me 😂
What does it mean Image datasets coloured and it has coloured with blue green rgb channels
Great explanation! Awesome Video!
Super excited for the CNN series 😁
Thank You so much, Sahil! 😀
Amazing explanation bro!!!
Wow Bro You explained it sooo Goood , I have a seminar presentation and i was searching youtube for this at last moment , your video helped me a lot Thank you!
Happy to help!
Amazing video !!! Keep up the great work
Thanks Achal!
THANKYOU SO MUCH!
3*3px leads to 9 parameters or 3*3*3 = 27 parameters (the trailing 3 is for RGB) ?
Could you plzzz teach me cnn from strach .. because after completely watched you videos I have certain doubts like ... How we initialize weight . How we resize it .. why we reshape it and also .. so many doubts ... Also I want to know about the syntax for all those ... Finally what I exactly want is 😢I want to write a code or train a model. Without ai help... End of the day I want answer it with my own brain cells ... Could you plzz help . Me ....
bro if possible share some notes too ,it would be helpful
This is a very under rated channel, needs more attention
Thank you :)
Bro , you said you upload data preprocessing video of house prediction dataset, please upload that soon , that will help us a lot
Hello ARES! Very sorry for that. I will be uploading that video, but it will take some time.
I have already created a written tutorial article on the topic you desire.
Here is the link for that: www.kaggle.com/jaimin09/simplest-way-to-reach-top-25-from-bottom-25
In this tutorial, I have explained the basics of data preprocessing and how to preprocess data for house price prediction.
Please do check it out. It's an article, but I have tried to explain everything properly.
I will create a video of the same in the future.
But until then, I hope this article can help you.
Let me know for anything else!
You are a rely good teacher. Thank you. You show a very good and structural understanding of what you are talking about. May I ask in which university are you studying your Bachelor?
And, if I may couple of technical questions:
1)
When you apply "cubic filter" (one filter for each rgb channel) and refer filter vales as weights, are they different for each channel thus, if filter size is 3*3 *3 I have 27 different weights? Or there are actualy 9 weights same for each channel?
2)It is not an obvious idea to use weights from fully connected to affect previous step - act as filters in the cnn layers. I mean same values could benefit as filters and harm the result as being used as weights in the FC, and vice versa - they can be helpful as weights but harmfull as filter values... Why do they do it such a way?
3) would you agree to the next idea: flattening different channels values of image by a filter cube to one single value when there are more then 3 channels (say Infra-Red) doesnt sound a good idea, since there are times that IR has totaly unique information that none of the rgb channels has. What do you think?
Thanks for the appreciation 🙂. It means a lot to me. I completed my bachelors from National Institute of Technology, Surat (India)
Regarding your questions:
1) Yes, if the filter size is 3*3*3, then there are 29 different weights.
2) Sorry, I did not fully understand your question
3) Yes, seems like a good idea. We can flatter the cube to a single channel filter.
hi!, why did the vertical edge detector capture some of the slanting and horizontal edges?
When you draw a straight line at a slant angle (best example is drawing a slamt line in paint), then you will see some horizontal bars at different heights. And due to the change in heights, there are small vertical edges. That is what the filter is capturing
@@CodingLane thanks for the awesome explanation 🔥🔥
@@arpit743 Your welcome
perfect
Great video
Thank you!
excellent video
Thank you!
you nailed it bru!
Thanks!!
I have one problem.
is every ML model has a neural network? If not please give me a summary of the ML Model with ANN & ML Model without ANN
Hello, we use Neural Network only in Deep Learning. Deep Learning is a subset of Machine Learning.
There are so many models which use variation of Neural Network. Example are Convolution Neural Network, Recurring Neural Network. There are many variations of CNN and RNN as well. Some names are RCNN, YOLO, LSTM, GRU, etc
And there are many models in ML which are not a part of DL. Some of them are: Naive Bayes, SVM, Random Forest, etc. These models do not use NN
@@CodingLane Thanks to your reply.
Does supervised & unsupervised always use NN? Can you mention any learning that refers to the same problem using ML with NN & ML without NN? Please. Then I can fully understand.
@@miniwin5791 all above mentioned models can be used for supervised learning. So, it is not necessary to use NN based model for supervised and unsupervised learning. You can use any.
You can apply any suitable model in an application, whether it may be NN based or not, it will work. But accuracy and performance will vary from model to model.
I have made a video on hand written digit recognition. You can check that out. In that, i have used ANN, but you can get almost the same performance using any other model which are not based on NN.
@@CodingLane Thanks Bro
unique concept !
Thanks
What are your qualifications ?
Hi Aadarsh, I can’t say that I am an expert in ML... but I like learning things and contributing the same. And I am still a student, completing my final year of bachelor’s degree.
CNN is ML or DL ?
CNN is a Part of DL, but DL is also a part of ML. You can checkout my video on ML vs DL vs AI. You will get better clarity
Can CNN implemented to a game ??
Lol🤣.
why not
i like you i support you