man your'e awesome . Even my MIT teachers fails to explains like you did. They just keep on saying blah blah blah didn't care whether anyone is understanding anything. Your'e a gem
Totally agree, I have tried doing the deep learning course by prof Mitesh Khapra of IIT M, he may have post doc, he may have very in depth knowledge, but when it comes to explain the difficult concept to "not so brilliant" students, he fails. Whereas Nitish sir explains it so easily that even an average student is able to understand.
00:01 Introducing the famous MNIST dataset for handwritten digit classification. 02:30 Handwritten digit classification using neural network architecture 05:26 The video discusses loading the famous MNIST dataset and understanding its structure. 08:30 Create a neural network architecture for digit classification 11:16 Setting up layers and architecture for image data in Sequential Model 14:03 Explanation of model architecture and parameter calculation 16:51 Discussed label coding and model compilation for training 19:28 Handwritten digit classification using ANN and the MNIST Dataset 22:09 Adding another layer and increasing model parameters 25:09 Using artificial neural network for multi-class classification problem 27:52 Handwritten digit classification using ANN and MNIST dataset
Hi Nitesh thank you for giving such lucid videos. I have seen other trainings including from Harvard or MIT or other ed tech platforms, but your videos are best to understand. It was your video which made concept of perceptron and MLP literally crystal clear. Thanks a ton sir . Alao taking this occasion to say Happy Guru purnima as you are virtual Guru for me in my journey of AI🙏
very well explained, Nitish it would be great if you could explain how 60000 image input is given to the first layer of ANN after converting to 1 D array, we have 728 pixel in each image, how they are connected to first layer
Sir, you ealier said at the start of playlist that we do not need feature engineering in Deep Learning, but we also need feature engineering to get better results in Deep Learning, infact in these two examples you done feature engineering to train Neural Networks. So feature engineering is required in Deep learning?
kya bhai, sir itna mehnat kar rahe aur tum abhi in sab chizo me hi fase ho , recommend it to as many people you can , make this guy popular because he deserves it
why does this not work from custom images(28,28) created from paint not predict properly, even after tweking the hyperparameters. i created digits [0,2,4,5,6,8,9] on paint wtih size(28,28). I got 1 out of 7 correctly (at this point it was just random) for x in [0,2,3,4,6,8,9]: img=cv2.imread(f'data/numbers/{x}.png')[:,:,0] img=np.array([img]) prediction = model.predict(img) print(f' result :{np.argmax(prediction)}') plt.imshow(img[0]) plt.show()
great videos. The playlist says 100 days, but there are only 76 videos. Does it cover all the topics or am I missing any video (i have not watched all the videos just want to make sure that I am referring the correct playlist or not) Appreciate if someone or CampusX can provide inputs
The video list is still not complete. He's still in the process of finishing it. So don't worry: 100 days is just an approximate number and it's currently ongoing
Namaste guruji, i trained my model and applied StandardScalar, somehow managed to make changes on model and now it gives 99.15% accuracy, now i want to test it on real world data, so should i reconvert new data to standard form or use serialisation? Using pickle is not good because if i update model or preprocessing pipeline, the pickled files might become outdated. Any suggestions how to do it seamlessly? Your useful insights and suggestions will be highly solicited. Thanking you in advance and await your reply.
Why we just took a prediction that the index with highest probability is the number, I mean why is it true what if our 3rd node is predicting for 8, or any i node predicts for a j value?
It's hard to express in words how good your teaching method is, and thank you for providing all these material for free.
man your'e awesome . Even my MIT teachers fails to explains like you did. They just keep on saying blah blah blah didn't care whether anyone is understanding anything. Your'e a gem
Your concept is crystal clear, your teaching style is superb. Hats off to you Sir.
Sir, your efforts have to be appreciated cause you wake up early in the morning and make videos for us. Thank You, Thank You So Much!
You are billions times better than my IIT professor
i agreeeeee
totally agree
Totally agree, I have tried doing the deep learning course by prof Mitesh Khapra of IIT M, he may have post doc, he may have very in depth knowledge, but when it comes to explain the difficult concept to "not so brilliant" students, he fails. Whereas Nitish sir explains it so easily that even an average student is able to understand.
Thank You please don't stop making videos your knowledge is adding a lot of values in our life
VERY GOOD ! 💪
you are totally right
link to colab is everything we need in description of video
amazing content sir I just wanted to learn ANN from you and here you upload it . thankyou sir
Your teaching is so addictive , i cant distract even i if i try , thank you :)
I thought I'm the only one who's addicted to sirs videos😅
why is this channel not having a million subscribers already?
28:20 ...Pichewali Chhipkaliii ko bhi concept smz aa gya sirrrr😂Itna Perfect kese ho skte ho aap...
MIND BLOWING Explanation🔥🔥🔥🔥
Thanks for this clear explanation.
Can't explain how ease it become after watching your video. Thank you so much for this playlist.
i did some optimization and got 98.6% accuracy.. Thank u sir for teaching us such wonderful things .
I really love your videos.I'm from Pakistan
finally getting good understanding on deep learning
thank your sir , bing watching your playlist like a netflix series
i recommend your channel to my every studnt
00:01 Introducing the famous MNIST dataset for handwritten digit classification.
02:30 Handwritten digit classification using neural network architecture
05:26 The video discusses loading the famous MNIST dataset and understanding its structure.
08:30 Create a neural network architecture for digit classification
11:16 Setting up layers and architecture for image data in Sequential Model
14:03 Explanation of model architecture and parameter calculation
16:51 Discussed label coding and model compilation for training
19:28 Handwritten digit classification using ANN and the MNIST Dataset
22:09 Adding another layer and increasing model parameters
25:09 Using artificial neural network for multi-class classification problem
27:52 Handwritten digit classification using ANN and MNIST dataset
binge watching your entire series
thanks for the great content,
Thank you Sir... Ab jake kafi cheeze clear hue :)
sir aap feedback ka toh pucha hi mat karo, ur always best. :)
mind blowing sir it , your teaching method is like makkhan
Sir BERT aur transformer models 100 days DP vale mein cover karoge ya NLP vale mein
Dono me daal denge
@@campusx-official sir topic nlp kab tak cover hoge.... Usko thode jaldi kardijiye. If possible
Hi Nitesh thank you for giving such lucid videos. I have seen other trainings including from Harvard or MIT or other ed tech platforms, but your videos are best to understand. It was your video which made concept of perceptron and MLP literally crystal clear. Thanks a ton sir .
Alao taking this occasion to say Happy Guru purnima as you are virtual Guru for me in my journey of AI🙏
Excellent sir.❤
hahhaha hum ny compile ni hum ny complile kia ha. love you i watched your every video .learn alot about deep learning and your big fan
very well explained, Nitish it would be great if you could explain how 60000 image input is given to the first layer of ANN after converting to 1 D array, we have 728 pixel in each image, how they are connected to first layer
Sir you took exactly 128 nodes in hidden layer how to choose to take that?
Please mention Day number for every video to keep track
Sir, you ealier said at the start of playlist that we do not need feature engineering in Deep Learning, but we also need feature engineering to get better results in Deep Learning, infact in these two examples you done feature engineering to train Neural Networks. So feature engineering is required in Deep learning?
thank you so much sir making this types of vdo
best yt resource for DS!
Nice sir, please make videos for deep learning
thank you sir for your effort .
nice content sir. really helping
@campusx please let me know why we used 128 node in hidden layer
Great video 🙏
Very informative video
Can we check here any other images except from the mnist data set ??
28:24 Lizard behind Nitish sir😳
sir, i don't want to like and share because everyone will come to know about this masterpiece
kya bhai, sir itna mehnat kar rahe aur tum abhi in sab chizo me hi fase ho , recommend it to as many people you can , make this guy popular because he deserves it
can anyone tell me how do I get the all the trained weights
Amazingggggg!!!!!!!!!!!!!!!!!!!!!!!!
13:07,Hi Bro,why you havent used flatten in churn prediction case
churn prediction is 1-d array, not 2-d array
Wah maza A Gya
very nice bhaiya
i am big fan of you
thank you sir
how I give a dataset of an experiment as a input and process the same whole programme...response please...
can any one explain.. hidden layer mai 128 nodes he kyu hai?
You can keep it any
One small doubt - Why we had to divide by 255 ? Why can't we use Standard Scaler to standardize the data.
We can't have negative values as pixels
why devide by 255 of x_train?
why does this not work from custom images(28,28) created from paint not predict properly, even after tweking the hyperparameters. i created digits [0,2,4,5,6,8,9] on paint wtih size(28,28). I got 1 out of 7 correctly (at this point it was just random)
for x in [0,2,3,4,6,8,9]:
img=cv2.imread(f'data/numbers/{x}.png')[:,:,0]
img=np.array([img])
prediction = model.predict(img)
print(f' result :{np.argmax(prediction)}')
plt.imshow(img[0])
plt.show()
what if images are MNIST-back-image, how to denoise them ??
Did we assume that the hidden layer will contain 128 nodes or did i miss something?
Yes, you are correct. We are only assuming there are 128 nodes.
how can i create a hand digit data..
easy to understand
Thanks for all your sessions. Do you plan to start a paid batch for Deep Learning in near future similar to the lines of Machine Learning batch ?
Hello sir, please make an video about validation_split. I have doubt about this
Where 255 coming from? 10:13
Sir kindly make a tutorials streamlit python
What about the color image digit recognition it will have 3 different RGB value unlike black and white all same value
Amazing sir😄
Sir what about back propagation
Will teach in upcoming videos
great videos. The playlist says 100 days, but there are only 76 videos. Does it cover all the topics or am I missing any video (i have not watched all the videos just want to make sure that I am referring the correct playlist or not) Appreciate if someone or CampusX can provide inputs
The video list is still not complete. He's still in the process of finishing it. So don't worry: 100 days is just an approximate number and it's currently ongoing
Shukria
Thanks 🙏
sir, model. predict(X_test) is giving us prob,but when I see the array its value is greater than 1 so how is argmax converting to digit?
why the image is colorful??as u are saying the image is in the range of 0 to 255
It's matolotlib color pallete
Namaste guruji, i trained my model and applied StandardScalar, somehow managed to make changes on model and now it gives 99.15% accuracy, now i want to test it on real world data, so should i reconvert new data to standard form or use serialisation? Using pickle is not good because if i update model or preprocessing pipeline, the pickled files might become outdated. Any suggestions how to do it seamlessly? Your useful insights and suggestions will be highly solicited. Thanking you in advance and await your reply.
Amazing
Hidden layer me 128 nodes kyun liya?
Bcz ye compulsory nh hai ap jitnen b leskty ho
Why we just took a prediction that the index with highest probability is the number, I mean why is it true what if our 3rd node is predicting for 8, or any i node predicts for a j value?
First Viewer and Liker
Thanks sir
finished watching
To run model and all ...deep learning is easy thae ML..
am i think right ..? (With keep in mind diffrence between ml and deep learning model)
best
you are doing the practical part before the theorey, that's why it is a bit difficult to understand, also the code section is also tricky.
Thanks
WHY 128 NODES IN THE FIRST HIDDEN LAYER????
You can select any number of nodes
Thanks you sorrrr
can we do the same project without using any extra library if yes then please make a video
can anyone talking a proper notes !? If yes then share with me please 🙏🏻
you seem to be a brother of @krish naik 😅🤣
Achcha hai but jyada samajh me nhi aa rha hai Nitish
nitish.campusx@gmail.com ek baar ispe explain karoge issue?
finished coding
hooray
speak english sir
Sir, your efforts have to be appreciated cause you wake up early in the morning and make videos for us. Thank You, Thank You So Much!