Vision Transformer explained in detail | ViTs

Understanding Vision Transformers: A Beginner-Friendly Guide:

In this video, I dive into Vision Transformers (ViTs) and break down the core concepts in a simple and easy-to-follow way. You’ll learn about:

Linear Projection: What it is and how it plays a role in transforming image patches.
Multihead Attention Layer: An explanation of query, key, and value, and how these components help the model focus on important information.
Key Concepts of Vision Transformers: From patch embedding to self-attention, you'll understand the basics and gain insight into how Vision Transformers work.
Whether you're new to transformers or looking to build a stronger foundation, this video is for you.

Make sure to like, subscribe, and comment if you found this helpful!

[soravsingla]

Your videos are always unique & highly knowledgeable. Thank you

[nagamanigonthina]

Excellent Explanation. Very well explained with basic concepts.

[layamahmoudi]

Thank you for the amazing video, it's absolutely perfect!

[vcarvewood]

You are excellent teacher. I'm in love in your voice since YOLOv8 tutorials. Attention to Aarohi is all we need.

[TruthOnly_jayshreeRam]

awesome, very nicely explained. Thanks Ma'am.

[AsthaPatidar-wt]

Please make a video for Convolution to Vision Transformer in detail.
And thanks for this video.

[bharatto]

Thankyou for explaining the videos very elaborately and clearly. But at some places it was too basic like RGB, would appreciate a timeline so that I can skip to the required part

[munimahmed]

Can you please explain DEiT model, this Vit explanasion is the best video on Vit I found on the internet. thanks a lot

[salmareang]

I have some confusion take one input image then how qkv are find ?

[madhavanu]

Transformers for remote sensing classification paper explain it ma'am...bcz you do it great and in easily understandable manner

[satvik]

43:20 You said that we do element wise addition of Patch representation and position embedding which means their dimension is same.
The patch representation is of length 768x1 and you also said the length of the position embedding vector is 512. How will you do the element wise addition. did you mean linear projected vector of eatch patch which has dimension of 512?

I learnt alot of stuff, thanks

[aryarushipathak]

Hello ma'am, can we use Vit and CNN to identify emotions from the face ? CNN for feature extraction and mtcnn for emotion labeling

[jynpogger]

God Please Protect My Teacher at all costs

[Mulugeta-cq]

Thank you for your Good work and can you make a video for ViTPose code too?

[CollegeOnline]

mam please please please please please please please create video on Gated Vision transformer as i am trying to use it in my research paper, but I am not able to find any literature regarding GVT. mam if you have any links to GVIT then kindly share it please

[ramchandhablani]

Mam, your video is very good, I have two questions, If there are 2 hidden layers, then there will be three matrices say W1, W2 and W3 for linear projection. The 2nd question is to train these weights and biases, we neet target vectors corresponding to each input vector. from where we will get those target vectors?

[nursami]

Hello ma'am, can you explain in more detail about encoder transformers such as normalization, multihead attention, softmax, MLP,
The video doesn't provide a detailed explanation about that, can you explain that in the next video?

[adityanjsg]

Arohi ji, possible for you to build a Model which is as good as gpt?
Though on limited data and scale..

[rahulhanot]

when will we get more video of this topic

[sreenalakhani]

pls make a video on video vision transformer also