Convolutional Neural Network in Matlab

[@awais2451985]

one of the best tutorials on implementation of CNNs. Thank you so much. God Bless you.

[@fadiliqbal8329]

Hey man, thank you so much for making these videos. Your English is amazing and your teaching style is brilliant. I am following your courses this week and it has been very helpful! Much love from Sri Lanka

[@mohamedabdallah400]

You are a genius, please don't stop, please continue, I am Arabian and I am not very good in English but I understand every thing from your pronunciation

[@Mba12509]

I dont know who you are but this is the one of Best lecture in the KZbin thank you bro go on May Allah bless you

[@NuruzzamanFaruquis]

Thank you Mohammed. These encouragements are what keep me going.

[@ibrahimmuhammed5332]

Quite impressive in everything....! Allah bless!

[@senakabandaraofficial]

Best tutorial on CNN!

[@QYL-dn2fo]

Thank you for your video！I learnt a lot from it！

[@gengli681]

Thanks for your tutorial. I follow the tutorial and learned a lot. And I find the accuracy is always if epochs is 3. But the filters are initialized randomly, so the accuracy should be a little different, I think.

[@NuruzzamanFaruquis]

I am glad that it helped you. The accuracy enhancement of ConvNet is an excellent research topic. You can work on different papers about accuracy enhancement in different conditions.

[@ashikeshan1529]

Very nice explanations and graphical presentation. Makes so much easier to understand. Thanks

[@NuruzzamanFaruquis]

I'm glad that you find it easy to understand.

[@jamesryanmofan855]

How to apply this code sir to your previous videos in Classification of fruit Using CNN

[@deepthakkar765]

where can we get the codes use in the examples? If feel some places in video it is not possible to type them?

[@melisaklc4149]

Could you please help me about these questions: How can implement this federated learning? How can i take gradient of the pics to implement federated learning?

[@nomadiccloud1529]

A very clear lecture. A great resource for new learners.

[@tanviralam4609]

Great video Mr.Nuruzzaman. Keep it up.

[@NuruzzamanFaruquis]

Thank you Tanvir :)

[@الشمستزورالمساء]

What happened in this code If I need to detect a watermark image ?? Thanks sir

[@RiteshMagreIT]

How to use convolution neural network for visual speech recognition?

[@NuruzzamanFaruquis]

An excellent question. I am saving our question as a topic. In upcoming lectures, I will try answer this question.

[@RiteshMagreIT]

@@NuruzzamanFaruquis Thank you

[@xingcui8680]

very nice ! hope future video have more about deeplearning!

[@NuruzzamanFaruquis]

Thank you. It will be a complete course on Deep learning starting from definition of machine learning (completed) all the way to single layer neural network (SGD, Batch and Mini-batch), multi-layer neural network, classification in multi-layer, Deep Learning (completed), ConvNet (completed). I will end the course by showing how to moderate Alex net to train CNN to recognize anything.

[@sirinelamine7025]

why did you choose "bsize=100" in MNISTConv.m? Thank you

[@anismaliki3478]

Very precise. Thank you very much. you saved me from the misery of my final year project T_T

[@ikke341]

Hi mister, thank you for your video. I want to ask how to make it in Gui of Matlab?

[@troghlem352]

Real and Honest thanks! You helped me a lot. I traduced this whole program to scilab, and this helped me a lot to understand how it works. I can only advise other readers to do so. I might post a link my code and some details here if I have the time.

[@venuyamalakonda8939]

Can you send me the link of your code

[@mandalamith9877]

can i get the matlab code please

[@ביתספרחרוץ-רומןגלגור]

thanks allot for your help and explanations. can u attached the matlab code please? thanks

[@ninavaschool6727]

HELLO , Can you explain what are these codes doing? d = zeros(2,1); d(sub2ind(size(d),D(k),1)) = 1; e = d - y ; delta = e; e5 = Wo'*delta; delta5 = (y5>0).*e5; e4 = W5'*delta5; e3 = reshape(e4,size(y3)); e2 = zeros(size(y2)); W3=ones(size(y2))/(2*2); and how can I aotained the result from these codes thank you

[@helohelo1681]

Yes

[@rakeshyadla6952]

bro how to get graph with this programme along images output

[@NuruzzamanFaruquis]

Would you please be more specific? If you would tell me which type of graph, I could help you.

[@12omgomgomg]

Is it anyhow possible to use CNN to automatically detect both purely and barely visible defects in an image? I am at this moment doing my Master's thesis and simply looking for the best algorithms to detect some defects on kind of blurry images (noisy a little bit, as it is a non-destructive testing using microwaves of dielectric structures). Thanks in advance!

[@NuruzzamanFaruquis]

You asked an excellent question. Yes, it is possible.

[@amruthaajithan3722]

really informative video sir.. Iam working on medical image security. sir please do a video of segmentation of images using CNN

[@عبدالوهابفوزى-ث8ط]

Great work bro, Is there any direct link to download the MatLab code?

[@snehalgaikwad6436]

hello sir, i have to train the characters using hopfield neural network. size of each character is variable as they are handwritten. how to do it??

[@GauravDhage]

Can you please explain training code

[@AakashKumar-dw8ne]

Thank uh sir, my question is how I classfied the retinal images using CNN to find the accuracy of images?? Please explain.

[@NuruzzamanFaruquis]

Would you please explain the question a bit. I will little confused here.

[@kudiratjimoh5267]

Great work! I really learn a lot. Thanks for sharing your knowledge.

[@giusepperandazzo91]

Best explanation out there! Love this video…new subscriber for u :)

[@hudaali6596]

Hi sir, Thank you for your video

[@SohamChakraborty42069]

Hello sir! Great lecture. I have a doubt. At 16:48, I cannot understand the filter = rot90(squeeze(filter),2) line. I read up the documentations of those functions but I still can't understand the reasoning behind doing these transformations.

[@mandyooi9963]

Hi! Thank you for this video. I'm currently doing with my final year project, which requires me to identify water puddles. Do you have any idea where can I find dataset of water puddle? Also, I'm aware that in order to identify water puddles from photos, I have to separate the pixels using image processing by colours.

[@NuruzzamanFaruquis]

Mandy Ooi, you asked a very good question. I am not sure whether there is any dataset of water puddle or not. However, preparing such dataset will not be that difficult. Here is how you can do it - 1) Record 20 videos of 5-10 minutes duration of water puddle from different angle, under different light intensity (morning, noon, afternoon), and in different environment, 2) Extract the frames from the video (now you have images) 3) Convert the images into gray-scale image 4) After that, resize the image using Matlab or OpenCV That's it. Your dataset will be ready. I will prepare a tutorial for you about it. I hope it will help.

[@mandyooi9963]

Awesome! Thank you so much, I appreciate it!!

[@NuruzzamanFaruquis]

You are welcome

[@NuruzzamanFaruquis]

This is data-set pre-processing phase. That means you will apply image processing to your data-set before training the CNN.

[@NuruzzamanFaruquis]

Mandy Ooi, I have uploaded a video on 'Face Recognition using Matlab'. There I showed how to prepare dataset and load dataset in Matlab. I hope it will help you in preparing and loading dataset in Matlab.

[@herdianto7618]

Thanks, very good you presentation

[@dr.ahlammajead4287]

Thanks very much..can you explain layer and option functions for cnn pleas???

[@NuruzzamanFaruquis]

I will try. Thank you for the comment.

[@techzila5379]

sir can you do matlab simulation of neural network for non linear system with taking some research paper?

[@lingualoom.X]

Hello, I am writing from Turkey. How to estimate weight with matlab opencv deep learning? How to calculate kilograms from photos?

[@sollybennet3387]

Sir, Thank you very much for the video. I have implemented your code on MNIST and it worked well. But I have a doubt in implementing the same for character data set. If I have character image of different size other than 28 * 28 ( say 60*60) and 46 classes then what are the corresponding changes I have to make? What should be the weight matrix ? Hidden layer of how many nodes ?

[@pravatbiswal7081]

wonderful explanation. keep on

[@vaishnavisaxena3824]

What is 'ieee-be' argument in magic variable? What does it do?

[@renukadevi1000]

Very good explanation on CNN using Matlab. I'm doing my research in human expression detection using deep learning(CNN). Kindly send me the Matlab codes for the above explained concepts. It would be great help to proceed my research with good understanding.

[@dawachyophel]

I do have a set of downloaded matlab codes for image super resolution and its perfectly running fine with existing datasets and models. But i want to create new model with different datasets and try to run with the existing codes. But i couldn't do so. It will be very helpful if you assist in creating model from new datasets in matlab

[@elifasloymak6136]

How can I do (deltax_1*W1)? or something like that.

[@negalema9780]

i need deep learning based antenna selection for massive MIMO matlab program can you help mee

[@ayoubelmj9319]

how i can solve this problem in testMNISTconv.m ?? Error in TestMNISTconv (line 21) [W1, W5, Wo] = MNISTconv(W1,W5,Wo,X,D);

[@eylulmood4830]

i have the same problem did you solve it plz t..

[@ayoubelmj9319]

Eylul Mood not yet , i can’t solve it , i was searching for a solution in the last 2 weeks and i got nothing !!

[@eylulmood4830]

@@ayoubelmj9319 thank you for answering

[@eylulmood4830]

@@ayoubelmj9319 I solved it in Conv function change this line y=zeros(20,20,numFilters);

[@ayoubelmj9319]

Eylul Mood i change It but i still have the same error, can you send me your code if you can ! [e-mail]

[@raniash3ban383]

thanks a lot this is more wonderful but i can not understand display_network function what it do? and how to write the comma in while loop

[@omidas1082]

sir i have error in display_network last line in if condition

[@haiderabbas5792]

Thank you so much, sir. Do you have an explanation about the detection of brain tumors using deep learning with CNN

[@HH-pu3ow]

Thank you for your video! Do you also have something on CNN in matlab with time-series data in stead of images?

[@NuruzzamanFaruquis]

Still now, there is no content on time series data. However, I will cover them in upcoming videos.

[@lalitbindra1087]

Thank you for the explanation. I am working on my final year project and wanted to know if cnn can be used to detect roi in thermal images.

[@miguelramirez4037]

God Bless you. Thank you for sharing your knowledege.

[@qw4316]

Dear sir , I got stuck in matlab CNN about the project ,Could u help me to solve it ? I can send u my data . Thanks again .

[@snehalgaikwad6436]

hello sir, If we have only dataset, no labels then which modifications we should do in the code?

[@eltonjohn7274]

Hi! is the value of each filters in convolutional layer corresponds to its weights? I already trained a network in matlab and i want to simulate the prediction programatically in java by using the trained weights and biases of the network. How can i achieve this?

[@NuruzzamanFaruquis]

I need to check if the '.mat' file type works in Java or not. I will reply once I will have got the answer.

[@abdulrahmankhalid8788]

first of all I really thank you for your valuable lessons... I just wonder if you can upload the Matlab codes so we can follow them step by step ... they are too long for such beginners like me... thanks again

[@NuruzzamanFaruquis]

Thank you for the appreciation. I will share the code once this course will be completed.

[@mahalakshmishanmugam2730]

Sir please tell that CNN whether work in Matlab 2015 version or not

[@veerapandiyanveerasamy2303]

Hi can u explain, how to do for data set rather than using images

[@marcelletchepgoua5481]

Thanks for the video it is a great one I have encountered a problem on the implementation of the program When I have the test script I got an error stating 'Error using reshape Product of known dimensions, 784, not divisible into total number of elements, 1. '

[@saraakhtar104]

Hello. I am getting an error message ' The EraseMode property is notnsupported and will error in future release'. Is there an alternate code i could use. Thanks

[@bienkichentusaohoa9324]

why do we use 20 convo 9x9 filter? where is it come out? Any rule for the filters?

[@NuruzzamanFaruquis]

You asked a very good question. It is an example. You can change the number of filters.

[@ravijaiswal2536]

Hello, nice explanation of CNN architecture. I want to know that, can we use CNN in structured dataset. Particularly CNN implement mostly in unstructured data.

[@dr.b.srinivas5505]

I have inspired for your explanation and video. I request you that, please do video brain tumor segmentation using CNN. thank you sir.

[@yellakc]

Great video tutorial. I really appreciate all the work you have done here. I have copied the code into matlab and I believe I’m working with the same dataset as you have. However, I had to change some small parts of the code (probably due to update to the matlab syntax handling)I’m getting less than 10% accuracy in 3 epochs unlike you (>90%). Is this normal? I couldn’t quite understand how you are updating the weights in the code because it seems you are adding them rather than w(k+1)=w(k) -rho*gradient for GD. I can see that you are using mini batch gradient descent with momentum but, for me momentum is: V(k+1) = beta*V(k) - (1-beta)*gradient and weight update changes to w(k+1) =w(k) - rho*V(k+1). Do you by any chance have the written code for testing purposes since I’m getting a different accuracy? Thank you for your time.

[@yellakc]

To be more clear: I have not changed your code before running but, if I were to change it I would use the equations I mentioned for momentum and gradient descent above. I’m really asking for clarification in this case. Thanks.

[@yellakc]

Sorry, ***V(k+1)=beta*V(k) + (1-beta)*gradient (not minus in between)***

[@NuruzzamanFaruquis]

I understand where the confusion is. I will try my best to prepare an explanation. Would you please email me your concern at [e-mail] so that I can replay with the explanation?

[@yellakc]

Nuruzzaman Faruqui Yes sir, I will email you right now. Thank you.

[@nazishiqbal1046]

@@yellakc from where i can get this code???

[@ENGMAM2012]

Thank you so much!

[@hirutabera1947]

how can we use convnet in plant disease identification?

[@NuruzzamanFaruquis]

Extract the features using convnet. Then identify based on the features.

[@jamsherali3316]

sir could you tell me.. can I use conventional neural network for prediction of state of health of battery.. please guide me

[@bfly2121]

Great video! The simple explanation of CNN at the beginning real solidified my understanding of how this works. In line 6 of "loadMNISTLables" what errors might occur? Im trying to run this code to get an idea of how it works but I'm getting the error code for that line when running it.

[@snb4799]

Sir thank u for everything You prepare perfect video , I learned so many thing and I am grateful to you for what you did.

[@imranuddin5526]

Great lesson, thumbs up

[@rashmiranjanrath8619]

i want to know how i get data from a emg signal

[@NuruzzamanFaruquis]

Would you please explain the question with a little more details?

[@yyren012]

Hi sir, Thank you for your video. I'm currently working with wood defection detection and classification using CNN. I already got 3 types of wood defection data and i'm wondering how do I write the coding for training? In this video, you didn't explain the training coding. Could you please explain the training code? Sincerely appreciate.

[@k.pritamdas2181]

Sir Please explain : What does 'D' in MnistConv.m represent? And the line: blist=1:bsize:(N-bsize+1) Is bsize the number of hidden layers?

[@vijinap1244]

Best tutorial for cnn in matlab Can u please explain how features are extracted from mri brain images for early diagnosis of alzheimer's disease

[@pushpag1572]

Can you send the complete code of this.?

[@ECEAnshuSingh]

Mnist dataset is a paid dataset?

[@benhamza643]

Hi! Thanks for this helpful video.I have a question, i have my database of color images with format jpg and with size different to 28*28,and i want to convert them to MNIST to train my network , what i have to do?(is it necessery to resize them to 28*28 or not?) And thank you so much

[@ismailbyibrine5176]

thanks for this video, how do we initialize weights ? is there any rules ??

[@muhammad.sanwal]

Sir can you pls explain the momentum command and epoch command used in your code?

[@BV2]

Hello, I think that your training algorithm at the end layer (dW0) needs to be multiplied by the partial derivate of softmax

[@hejran9148]

hi, can I implement this code on 1D-audio ?what is the change is require?

[@justinjose3067]

An extremely helpful video for a beginner like me!.. Can you also upload those matlab codes ?

[@NuruzzamanFaruquis]

Thank you for your compliment. I am working on uploading the codes. Once it will be done, you will find it in my website.

[@studywithme8741]

Did you upload the code.?

[@mohamedelmekawy-rw8vz]

@@studywithme8741 You can find codes here at this link ( github.com/Apress/matlab-deep-learning ) in ch.6.

[@FedePayes]

@@mohamedelmekawy-rw8vz thank you so much

[@hunter7ize]

Thanks for the ingisht sir, I have a question. How do I train the neural network to identify colors?

[@NuruzzamanFaruquis]

That's a very nice question. We use gray-scale image to train neural network. To identify color, you can split the images into channels. Based on the average value of each channel, you can calculate the intensity of the color.

[@hunter7ize]

@@NuruzzamanFaruquis do you have tutorials on this one sir?

[@azmanman3454]

Thanks sir for an excellence presentation , is CNN suitable in data prediction ? or its just only for image recognition

[@xdgamestudios2660]

Hi!! thank you so much for this video!! I have a question though. I'm trying to extract features from 1D array signal, the source is a EMG signal so I only have microvolts over time. the input arrays are 1X2000 arrays and the output is 1X2 (two classifications) "valid" or "invalid". Would I need to change the program a lot to make it work?!? in theory a 1D array could be treated as a small (very small) image with 1 pixel height and 2000 width. but then how about the filters... they couldn't be 2D arrays either, am I wrong?! also could you upload the source code ;) thank you

[@NuruzzamanFaruquis]

Convolutional Neural Networks are for image recognition. What you need to do is 1) generate numeric values corresponding to your signal 2) use those numeric value as input of a neural network. 3) train the neural network using supervised learning method You have two classes (yes or no), means it is a binary classification, a very simple one. I have uploaded a lecture on 'Deep Learning' which will put you in the right direction. The codes are given there as well. I hope your problem will be solved.

[@xdgamestudios2660]

Well so CNN is not what I should be looking for then!? 1) I already have the signals (they are 1D matrix with 2000 length) 2) now I must use this values as input, right? (meaning, I have 2000 inputs and want one output 1/0, 'yes' or 'no') 3) training is my problem (not quite sure I really understand what i must do. I have 500 signals (1X2000 matrixes) and I've already classified them all so I can verify if the output network value is right or wrong. I was going to use 450 as training and 50 as test. not quite sure I know how to train a NN though. I'll watch your DL video :)

[@NuruzzamanFaruquis]

Don't worry. In the architecture of the Deep Neural Network (I explained in the lecture), use appropriate number of input nodes ( 20000 I guess in your case) and output nodes ( two nodes). Then train and save the network (has been shown as well). That's it. Now every time you give the input, the saved network will give the true or false result. (Although single layer neural network is enough for your project, however, I haven't made any tutorial on Single Layer Neural Network yet)

[@bogdanelefteriu6945]

@@xdgamestudios2660 I would suggest looking into Euclidean Distance Matrices for this one. They can help you transform 1D signals into a 2D heat map image.

[@gulrukh1386]

Respected Sir, May I ask about your code that what is used for the following code-lines, please? dW1 = dW1 / bsize; dW5 = dW5 / bsize; dW0 = dW0 / bsize; Secondly, also requested, for the following lines as well:- W1 = 1e-2*randn([9 9 20]); W5 = (2*rand(100, 2000)-1)*sqrt(6)/sqrt(360+2000); W0 = (2*rand(10, 100)-1)*sqrt(6)/sqrt(10+100); Description of W1, W5 and W0 alongwith dW1, dW5 and dW0. thanking you in anticipation. best Regards gul Rukh Khan

[@mamoonanisar2035]

I m learning CNN in these days .using Pytorch.please make a video on it.Also my Can CNN use for tumor "segmentation ".Highlighting the tumor from image is required task.

[@AlbertWang1]

where to download the code?

[@abaasalbasha4575]

Hi, thanks for this explain wonderful for video I have a research that is colorization of grayscale images .Do you have any idea about it

[@NuruzzamanFaruquis]

Yes, I do. I will upload the video of grayscale to color image conversion shortly. You can look into previous comment. There is another guy, Eman Saleem, working on the same area . Both of you will get a notification when I will upload the lecture. Thank you.

[@abaasalbasha4575]

Nuruzzaman Faruqui Thank you very match

[@raniash3ban383]

hello thank you i need to ask what for loop of epoch do

[@pragyansahoo2967]

hello sir,, can CNN be used for classification of medical data? I mean to classify wheather the patient has some disease or not? how?

[@rahulsinhchauhan5503]

Is it possible by image processing by CNN 🤔..Have u any code

[@abdullahabdurahim8845]

thank you , it is best course in deep learning , but the big equation is (how use CNN?) , How can I benefit from this network?

[@robinschafer597]

Tank you for the whole playlist on this topic. Awesome work!!! Did you publish your scripts and functions for interactive learning?

[@unnamedunnamed7212]

Hi! Thanks alot for your videos I'm currently doing with my final year project, which requires me to identify edge in seismic data using DNN in matlab. Do you have any resources that can help me?

[@NuruzzamanFaruquis]

Would you please be a little more specific?

[@unnamedunnamed7212]

​@@NuruzzamanFaruquis Thank you for replying , I need DNN matlab source code help me to train my model that can detect the edge in seismic or public images. If there is no DNN source code can you give me CNN source code to detect in public images.

[@manfredkremer2103]

Thank you for the nice tutorial, but this is pretty complicated and very fast. Where can I find the complete code, then it would be easier for me to follow!

[@SIKANDERSINGH-nb7rz]

can you make a code for islanding using deep learning for current.

[@rameshjaganathan6444]

Thank very much for your excellent explanation about CNN. I'm doing my research on image processing. Can u suggest me how to use for image segmentation.

[@ajaybazilissac4557]

can u implement alexnet ?? is there is any difference???

[@emmanuelkwekumensah3566]

hello and thanks for this informative tutorial. Please is there a link to access the code??? Would like to implement it myself.

[@NuruzzamanFaruquis]

Thank you. There are multiple functions. I couldn't paste them in the description. However, as I am getting numerous requests for the code, I will try to come up with a solution to provide the code. In the meantime, you can take the code from the last section of this video and type it in MATLAB.

[@emmanuelkwekumensah3566]

thank you. however i would recommend you set up a github account to support your channel.

[@NuruzzamanFaruquis]

Thank you for the suggestion. I will look into the possibilities.