Image Classification using Vision Transformer (ViT) in TensorFlow

In this video, we will build a flower image classification using Vision Transformer, which is implemented from scratch in the TensorFlow framework using the Keras API.

Vision transformer (ViT) is a transformer-based architecture used in the field of computer vision, it is directly inspired by the use of Transformers in NLP tasks.

Timeline:
0:00:00 - Introduction
0:00:31 - Dataset Explanation (Flower Images Dataset)
0:01:00 - Importing, Seeding, Dataset and others
0:37:22 - Implementing the Vision Transformer (ViT)
1:04:30 - Training the Vision Transformer (ViT)
1:08:49 - Testing the Vision Transformer (ViT)
1:13:57 - Ending - SUBSCRIBE

Support:

Follow Me:

[teetanrobotics]

bro please put all the videos in playlists. So it is organised and the audience can binge watch. Thanks for the amazing content. keep it going.

[hamedshokripour]

Very accurate and useful tutorial. Thank you very much

[aditisengupta]

Thank you, sir. It was a very helpful video. :)

[ahpacific]

Hi - this was very helpful but can you please provide the code to inference on a single image - not a evaluate a whole set? I normally use when training with traditional classifiers but in this case this does not work. I don't know exactly what preprocessing to do before using model.predict() on the image I wish to inference with.

[nehal]

Very good[VIT].. request make a detailed video on GAn /condition GAN and their
implementation

[shushankyadav]

Can you please show how we can evaluate the model based on performance matrix such as precision recall f1 score ETC.

[waqarahmed]

how can i modify this code for multilabel image classification?

[kirtishyamsukha]

can you tell what is the accuracy and other performance matrix after the 500 epochs, also please show how to print the predicted classes.

[ashmalvayani]

can you show us how to make predictions on run time with a specific image by displaying it's output as well on screen using the .h5 weights?

[aryam]

Thanks a lot for the detailed video on image classification using Vit. Is there any way to find out or visualize the attention map as well in tensorflow.

[rama_gpubhuyan]

Thank you for such detailed video.
Just one question; if the model size is such huge, where exactly we will use it. And how to maintain a competent inferencing time CNN based models.
what is its practical usage?
Thank you

[sithibanu]

Can you explain the SAM model in tensorflow?

[ashishrana]

I don't understand how the transformed input shape which is (64, 1875) to (256, 1875). Please help me out to understand it?

[fayezalhussein]

amazing series, dear i am struggle with replace the dataset with another, could you explian how i could use another dataset, please

[kenand]

"x = MultiHeadAttention( num_heads=cf["num_heads"], key_dim=cf["hidden_dim"] )(x, x)" What does (x, x) mean here? Can anyone explain?

[loveofmylifesoumyarashmi]

which gpu is being used here? and cant we do the same by using Jupiter?

[RahulDogra-sd]

Can I use this for document classification for files like pdfs, text files, docx?

[anitachoudhari]

Hey! I am getting error while running test program...please do reply kal dikhana hai project😅

[mariaachary]

can u share you swim transformer video also with visualization

[ahmedchaoukichami]

can i use transformers in regression problem