Working with Images (MNIST) | PyTorch Images and Logistic Regression | Model Training and Validation

preview_player
Показать описание

💪 Working with images from the MNIST dataset, Training and validation dataset creation
⚙ Model training, evaluation, sample predictions and more simplified at a beginner level

Code and Resources:

In this tutorial, we'll use our existing knowledge of PyTorch and linear regression to solve a very different kind of problem: image classification. We'll use the famous MNIST Handwritten Digits Database as our training dataset. It consists of 28px by 28px grayscale images of handwritten digits (0 to 9) and labels for each image indicating which digit it represents.

Time Breaks
00:00 Introduction
04:14 Working with Images and Linear Regression
14:52 Training and Validation Datasets
21:27 Defining our Model
46:20 Evaluation Metric & Loss Function
57:55 Training our model
1:27:48 Testing, Saving and Loading the Model
1:45:15 Assignment 2 - Train your First Model
1:56:39 Course Overview
1:58:24 What to do Next?
2:00:16 Jovian Data Science Mentorship Program

Topics covered in this video:
⌨️ Working with images from the MNIST dataset
⌨️ Training and validation dataset creation
⌨️ Softmax function and categorical cross-entropy loss
⌨️ Model training, evaluation, and sample predictions

Deep Learning with PyTorch: Zero to GANs is a beginner-friendly online course offering a practical and coding-focused introduction to deep learning using the PyTorch framework.

This course is taught by Aakash N S, co-founder & CEO of Jovian - a platform for sharing, showcasing and collaborating on data science projects online

--
Рекомендации по теме
Комментарии
Автор

Whole list, structure, your website with its resources is simply amazing! Great job and lots of success to you!

davidlearnforus
Автор

Excellent flow and a very structured teaching. Thank you so much.

ashish
Автор

Hello, Jovian. The course seems to be phenomenal and I am really enjoying it so far.

I have a question. At around 1:30:00, why are you using the unsqueeze function to add an additional dimension to the batch, instead of just reshaping the input image like we did in the training stage? I did the reshaping method and got the same results, so I am just curious about it.

SillyLittleMe
Автор

Your explanation is too good sir🙏Best course

Karthik-kt
Автор

In the MnistModel class, why do we call forward using "Self(Images)". Why dont we just call it as Self.forward(Images)?

Also, why's it the same when after instantiating too?

Why
model = MnistModel()
outputs = mode(Images)


Instead of
model = MnistModel()
outputs = mode.forward(Images) ?

everythingaccount
Автор

Your videos are really great. You can learn a lot. But many of your video tutorials have no subtitles, which brings a lot of trouble to people from non English speaking countries.

xuqi
Автор

this is great work by jovian to provide free series on pytorch so what will be your next course and when? thanks for such a great course

anuragshrivastava
Автор

@1:29
Single Image size = [1, 28, 28]
As mentioned, model takes a batch of size -> [128, 1, 28, 28] (128 images, each of size 1, 28, 28)
We flatten it using reshape [-1, 784]
So, my question is even if we send a single image or a batch, does this matter because anyways the model will flatten the image?
Single image reshape -> [1, 28, 28] -> reshape(-1, 784) -> [1, 784]
Batch reshape -> [128, 1, 28, 28] -> reshape(-1, 784) -> [128, 784]
Can somebody verify this or am i missing something?

switchwithSagar
Автор

the second cell about downloading mnist is not executing in colabs showing an http error plz help me out in logistic regession

rockstarsai
Автор

@36:51, how did weight and bias get populated even when they were not defined in the MnistModel class?

tj
Автор

Can someone explain why wasnt sigmoid function used here

Om-idqr
Автор

when you mention only one channel for img_tensor what does it mean and why is only 0 present as one channel why cant we take 1 instead of 0 as its a gray scale image even 1 should be present right ? is it because the indexing starts from 0 instead 1 is that why we put 'print(img_tensor[0, 15:20, 15:20])' instead of ' print(img_tensor[1, 15:20, 15:20]) ' or is it because of some other reason

when i remove 0 and add 1
print(img_tensor[1, 15:20, 15:20])

i get this error " index 1 is out of bounds for dimension 0 with size 1 "

flowerboy_
Автор

Nice video!!! But without subtitles😂😂😂

guaiguaixxx
Автор

hi. could you please explain how come the code works even though you didn't use softmax in Model() class?

sajalhsn
Автор

Why is logistic regression a Linear Model?

arpit
Автор

very bad.
your online session with free camp code are goated, Worst I can say this playlist

pml
visit shbcf.ru