CNN is best for any set of images to classify it. I suggest you to prefer the datasets which consists of more images.

Dear Murana Awad ,

Finding the right dataset while researching for machine learning or data science projects is a quite difficult task. And, to build accurate models, you need a huge amount of data. But don’t worry, there are many researchers, organizations, and individuals who have shared their work and we can use their datasets in our projects. In this article, we will discuss more than 70 machine learning datasets that you can use to build your next data science project.

https://data-flair.training/blogs/machine-learning-datasets/

Regards,

Shafagat

https://www.kaggle.com/search?q=cnn+tag%3A%22cnn%22

There are various datasets that you can leverage for applying convolutional neural networks. Here are three popular datasets:

• MNIST
• CIFAR-10
• ImageNet

There are different datasets that you can used with CNN, I send you link that you can find various datasets to use it in your project...

https://www.kaggle.com/search?q=cnn+tag%3A%22cnn%22

You can find a huge amount of datasets here :

https://paperswithcode.com/datasets

You can find all types of datasets in the below link:

https://www.tensorflow.org/datasets/catalog/overview

Did you have try CIFAR-10 or CIFAR-100?

It is very hard to find a proper dataset that will give decent training and testing/validation accuracy. There are several good public datasets present in the internet and you can always ask the author of a particular dataset published in his or her work. If you are going by trial and error (most of us do that!!!) then you have to go through a lot of datasets and also you have to tune your model accordingly. All the best.

It i import to eveluate the acquisition protocol of the different database received from the previus message. The quality of the acqusition and the technical parameter determines the quality of the final model. See 1. Physica Medica 87(2021)115–122 and 1. EMJ Radiol. 2020;1[1]:26-28, European Radiology Experimental (2020) 4:62

Dear Murana Awad

You can check these datasets.

• MNIST (9 classes and 60,000 images)
• CIFAR-10 (total of 60,000 color images in 10 classes)
• CIFAR-100 (100 classes and 600 color images per class)
• ImageNet (1000 classes and 1.2 million images)
• Stanford Cars (6,185 images of 196 classes of cars)
• CUB-200-2011 (Number of categories: 200; Number of images: 11,788)
• FGVC Aircraft (10,200 images of aircraft, with 100 images for each of 102 different aircraft model variants)
• Caltech-101 (9,146 images, split between 101 different object categories,)

Dear Murana Awad :

You may use convolutional neural networks to a variety of datasets.

Among the most commonly used datasets are the following:

MNIST \ Caltech-101 \ sCIFAR-10 \ ImageNet \ Stanford Cars

Regards;

Ehtisham

it really depends on the task and field (medical, other): for regression/classification MNIST, for segmentation the BRATS which are also 3D medical images, see Preprint Brain Tumor Segmentation and Survival Prediction using Autom...

Dear Murana,

It depends on your task, whether it is "Segmentation", "Classification", "Object Detection" etc.

For anyone of the above problems, there are tons of datasets. The "MNIST" and "CIFAR10" are the most popular for the classification task.

I recommend having a look at the "Kaggle" website in this regard. You can find really good datasets there.

And another important thing is that you should choose your dataset considering your hardware capacity.

I hope it helps

https://www.analyticsvidhya.com/blog/2020/02/learn-image-classification-cnn-convolutional-neural-networks-3-datasets/

https://www.kaggle.com

https://physionet.org

If you use the transfer learning technique, you must choose the largest data set