L-5 | Image Classification Using ResNet-18 and Pytorch

In this tutorial, you will embark on a journey through the world of image classification using the powerful ResNet18 model with PyTorch. We will guide you step-by-step on how to train the model on your dataset, capturing the nuances and features essential for accurate classification. Following training, you'll learn the critical skill of saving your model, enabling you to preserve and utilize your trained model in future tasks. But the learning doesn't stop there! You'll also master the art of loading the model back into your Python environment, a key step for deploying models or continuing training. Finally, you will apply the trained model to new data, testing its ability to classify images accurately. By the end of this tutorial, you'll be well-versed in handling the ResNet18 model for your image classification tasks using PyTorch.
Github: github.com/Aar...
Dataset: github.com/Aar...
For queries: You can comment in comment section or you can email me at [e-mail]
#computervision #deeplearning #pytorch #imageclassification #imageclassification

[@navidnazem4494]

Shouldn't we redefine FC to 2 classes?

[@benoitd94]

hi good tuto. One question i try to create my own dataset. You don't tell in the video if the label process is required ? with labelimg for example ? thx

[@CodeWithAarohi]

For Image classification task, You don't need to annotate the images. Just create separate folder for each class and then put related images.

[@НиколайНовичков-е1э]

Thank you, Aarohi! :)

[@CodeWithAarohi]

Welcome :)

[@tilkesh]

Thank you very much

[@CodeWithAarohi]

You are welcome

[@soravsingla8782]

What an amazing video

[@CodeWithAarohi]

Glad it is useful!

[@salihamirza456]

hey i have a question that,as we have copied the weights and biases from the loaded model to the new model, then shouldn't we use new_model for performing inference as well?

[@CodeWithAarohi]

Yes

[@therock8815]

plz provided the updated code of vgg-16/ resnet50 or any that resolved the version error thanks

[@Sunil-ez1hx]

Awesome

[@CodeWithAarohi]

Thanks!

[@karthickkuduva9819]

Mam. Freezing all layers except the final classification layer is called transfer learning. And customized with our own dataset so it's also fine tuned model. The way I understand is correct ?

[@CodeWithAarohi]

Yes, your understanding is correct!

[@youtubeaccount8613]

this is so so so helpful! i have been stuck for weeks and this is amazing!

[@CodeWithAarohi]

Glad it helped!

[@AlAmin-xy5ff]

Mam make a video on YOLOv8 model customization or change layer

[@CodeWithAarohi]

I will try

[@rafeeda3580]

Well explained.❤❤❤ Any videos of medical image classification available

[@CodeWithAarohi]

Not yet!

[@bilkisuismail6096]

Thank you ma, but how did you get mean and standard deviation?

[@dobijr7900]

they are gotten from imagenet after working on millions of images

[@huyngo8567]

Your link is broken

[@CodeWithAarohi]

github.com/AarohiSingla/Image-Classification-Using-Pytorch

[@ABHINAYYADAV-j2h]

it is giving error I am taking 11 classes for this for emoji prediction

[@ABHINAYYADAV-j2h]

from PIL import Image from torchvision import transforms # Load and preprocess the unseen image image_path = '/content/test.png' # Replace with the path to your image image = Image.open(image_path) preprocess = transforms.Compose([ transforms.Resize(256), transforms.CenterCrop(224), transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) ]) input_tensor = preprocess(image) input_batch = input_tensor.unsqueeze(0) # Add a batch dimension

[@ABHINAYYADAV-j2h]

RuntimeError: The size of tensor a (4) must match the size of tensor b (3) at non-singleton dimension 0

[@abrarislam6509]

Are you married?

[@CodeWithAarohi]

Yes

[@ashimasingla103]

Perfect content

[@pifordtechnologiespvtltd5698]

Nice video

[@CodeWithAarohi]

Thanks