Convolutional Neural Networks : TensorFlow 2022
Vložit
- čas přidán 25. 08. 2024
- I livestream code and explain everything you could want to know about Convolutional Neural Networks using TensorFlow.
Convolutional Neural Networks can be used to train computers to recognize objects in images. This allows us to recognize objects or people in images or video. It is helps automate the process of finding cancerous tumors. It is also used to implement driverless car technology and more.
Code is Here : github.com/der...
In this video I'll cover all of the following Live :
✳️ Uploading Data to Google Colab
✳️ Gathering / Preparing Images
✳️ Normalizing
✳️ Convolution
✳️ Activation
✳️ Pooling
✳️ Optimizing Performance
✳️ Minimizing Overfitting
✳️ Augmenting Images
✳️ Saving & Loading Models
MY UDEMY COURSES ARE 87.5% OFF TIL June 28th ($9.99)
www.udemy.com/
➡️ Python Data Science Series for $9.99 : Highest Rated & Largest Python Udemy Course + 56 Hrs + 200 Videos + Data Science bit.ly/Master_...
➡️ New C++ Programming Bootcamp Series for $9.99 : Over 23 Hrs + 53 Videos + Quizzes + Graded Assignments + New Videos Every Month bit.ly/C_Cours...
MY UDEMY COURSES ARE 87.5% OFF TIL June 5th ($9.99)
www.udemy.com/
➡ Python Data Science Series for $9.99 : Highest Rated & Largest Python Udemy Course + 56 Hrs + 200 Videos + Data Science bit.ly/Master_Python_32
➡ New C++ Programming Bootcamp Series for $9.99 : Over 23 Hrs + 53 Videos + Quizzes + Graded Assignments + New Videos Every Month bit.ly/C_Course_32
Well, Hello Derek and welcome to my comment :) Just here to say Derek Banas was a strong foundation in my early days dabbling in programming. Whatever I know in Python, HTML, CSS and Javascript was first learnt from Derek Banas.
Thank you Derek! Amazing tutorial series. Hope you will continue it.
Thank you very much :)
Hi Derek. Thanks for these awesome TF tutorials! By the way, just discovered that the CNN explainer site @1:11:06 isn't broke. You just need to configure the parameters like "Stride, Input, kernel Size" so that it will not show red boxes. Red boxes means that certain parameter is not compatible to other parameter.
Great content! Loving this tutorial series!
Thank you :) I'm happy I could help
Great tutorial - thank you Derek!
Thank you for taking the time to tell me :) I appreciate it!
Man, I'm pretty sure you must have a separate, detachable brain to store all your knowledge. Thank you for sharing it
That's funny. There is a lot of random things bouncing around in my brain. I'm happy you enjoyed the video.
Derek!
I remember your Unity series with these awesome clones of very popular games!
Zelda clone was just BLAST!
Where are those razors? ;]
How are you even doing Derek?
Do you live well?
Everything alright?
Maaan, I wish we can drink vodka someday together ;]
Thanks a lot for your time, I really appreciate it!
You're the best!
Thank you :) I'm happy you enjoyed those videos. I'll be teaching my Daughter how to make games this summer, so it is probably the time to make new video game tutorials. Thanks for taking the time to say hi! I greatly appreciate it
Thank you Derek! Btw, you no longer have to use the drive.mount() cell and copy the security code to connect to Google Drive. You can just click on the Drive icon on the Files navigator.
Dear Derek,
A friend of mine and I watched all your machine learning videos and are trying to solve similar problems for what you are talking about in each video.
Specifically, for this video we tried to train a model to classify cats and dogs, but sadly we weren't able to achieve great results (5 of his 6 cat pictures were classified as dogs) even though our model had ~83% accuracy and we tried every step to improve our model. The pictures we tested with look fairly similar to the training and validation data pictures, so we are confused why the results are so pro-dog.
Sincerely, we are in tears.
That's ok this is a learning process. In the video after this I show how to improve the classification results and also introduce transfer learning which dramatically improves results czcams.com/video/wJWtZq6f-60/video.html In my next video I'll improve on that by covering feature tuning. By the end you'll be able to classify images, text, etc. with unbelieveable accuracy.
@@derekbanas This is quite a *learning* process, indeed. Excellent pun, Derek. I have already watched the video you linked and we also applied to download a well-trained model, but as it turned out we used the wrong activation function. Now all of his cats truly are cats and we still are in tears, but in tears of joy. Thank you for your videos and also your reply (most of all your dedication), they are informative and fun to watch :) We will definetely tag along.
Have a nice day!
What is the actual Syllabus to learn CNN I mean where I need to start learning ?
How could one even out the amount of images simply and randomly of different classes, if I have x amount of one images, y amount of images of second class and z amount of third images.
It would be awesome to see this model in action with a real small robot that recognizes and goes towards the hotdog over a pizza slice! Would something like that be easily done with Tensorflow? To make it eat it would be very difficult though. And a tiny bit wrong!
PS : thanks for your tuts, always great and interesting videos that I'm jealous I cannot do myself lol!
Thank you :) Yes you absolutely can do that and I plan on covering how to in future videos. I'll use TensorFlow Lite to deploy my models in many internet of things applications.
Hi Derek, thank you for your tutorial.
I am not sure whether you will be monitoring the chats now. But I have a doubt when I followed your instruction and I am a beginner. During the training of the model I get the following error. Not sure whether I have made any mistakes somewhere.
AttributeError: 'NoneType' object has no attribute 'items'
However removing the validation steps (validation_steps = len(validation_data_norm)) and allowing the model to choose the suitable steps gets rid of this error but my accuracy is lower. I even tried with different number of validation_steps but it all gives the same error in some form or another. Further, I have got the same exact error during my other CNN models. Any idea why this occurs or is it any fault from my side. I am using jupyter notebook. Anyone who has any idea regarding this, your help will be appreciated.
here the problem was the validation_steps were giving a value about 16.5, so I changed it to validation_steps = num_val_images // batch_size and the code runs without error. But this gives a warning that dataset ran out of data before completing the specified number of steps per epoch. To fix this, I used the .repeat() function when preparing your datasets to ensure they provide sufficient (infinite) data for the specified number of steps. I am not sure whether this is the way to fix this error, but now my accuracy improved.
Hi, could you explain how to change the code if you are not working with binary output? It is really clear how it works with binary cases but you could say in the videos "if it is not binary do this".?
I cover that in this video czcams.com/video/wJWtZq6f-60/video.html
thanks Derek! awesome video! Do you recommend to go through your Udemy course, in case it has some stuff which is not here. Also would highly appreciate if you can give any homework or projects to do apart from these 6, or any sample question websites, which will help with getting ML jobs
Thank you :) My Udemy course is basically everything imaginable smashed into one course. So rather than saying learn MySQL some place else, it also includes MySQL. It teaches all the math and basically anything else related to Python and data science. I add additional videos every month and soon I'll make exclusive machine learning videos. I'm very proud of it. The only negative reviews I ever get is that it is too big. That is the idea. I wanted to make one course that covers everything.
Hi Derek, thanks for this tutorial! Do I understand correctly, that here, we are dealing with classification rather than object detection?
Happy you liked it. Yes we are looking at the images, defining shapes and then based on those shapes we decide if something is a pizza or a hotdog. This is classification. I'll cover how to find multiple objects in one image, or video soon. I also plan on making IOT projects where I'll train a robot to see an interact with its environment. I'm trying to train one to shoot little basketballs right now
@@derekbanas Thank you for the kind reply, Derek! Looking forward to the basketball-robot ("What is my purpose?" - "Playing basketball.").
This is only for two images but what if there are two images then how do we use 0 to 1 . I didn't quite get it
Watch this video for more czcams.com/users/livewJWtZq6f-60
@@derekbanas amazing
Im 22 couldnt go to college afraid its to late for me
Few day ago i found out about Anvil. It makes very easy to create web apps with only python. Just drag and drop components. Skipping html and css. Can you show us how to connect our ML models to the web app?
I have been looking at it but haven't had time to test it out. I would love for there to finally be a Python solution for app development!