Скачать или смотреть How to Use the Cifar-10 Dataset with tfds.load() in TensorFlow

Discover the best practices for loading and using the `Cifar-10` dataset using `tfds.load()` in TensorFlow. Simplify your deep learning projects with our step-by-step guide!
---
This video is based on the question https://stackoverflow.com/q/67635660/ asked by the user 'Lleims' ( https://stackoverflow.com/u/8551424/ ) and on the answer https://stackoverflow.com/a/67636211/ provided by the user 'Frightera' ( https://stackoverflow.com/u/13726668/ ) at 'Stack Overflow' website. Thanks to these great users and Stackexchange community for their contributions.

Visit these links for original content and any more details, such as alternate solutions, latest updates/developments on topic, comments, revision history etc. For example, the original title of the Question was: Using Cifar-10 dataset from tfds.load() correctly

Also, Content (except music) licensed under CC BY-SA https://meta.stackexchange.com/help/l...
The original Question post is licensed under the 'CC BY-SA 4.0' ( https://creativecommons.org/licenses/... ) license, and the original Answer post is licensed under the 'CC BY-SA 4.0' ( https://creativecommons.org/licenses/... ) license.

If anything seems off to you, please feel free to write me at vlogize [AT] gmail [DOT] com.
---
Loading the Cifar-10 Dataset Using tfds.load()

If you’re diving into the world of convolutional neural networks (CNNs) and want to practice on a well-known dataset, the Cifar-10 dataset is a fantastic choice. This dataset is widely used for benchmarking image recognition systems and contains 60,000 32x32 color images in 10 different classes. Though loading the dataset using simple functions like datasets.cifar10.load_data() is straightforward, many developers prefer using tfds.load() for added features and flexibility.

In this post, we'll walk you through the steps necessary to correctly leverage the Cifar-10 dataset using tfds.load() from the TensorFlow Datasets library.

The Problem: Loading Cifar-10 with tfds.load()

You might find it challenging to load the dataset correctly using tfds.load(). For example, you want to obtain both train and test datasets, and ensure they are in a format that is easy to work with. Here’s a snippet of code showing the initial attempt to load the dataset:

[[See Video to Reveal this Text or Code Snippet]]

From there, you may run into issues asserting the types of the loaded datasets, and attempting to extract images and labels. Here’s what that might look like:

[[See Video to Reveal this Text or Code Snippet]]

The problem occurs because you are not using the correct parameters to extract the images and labels.

The Solution: Properly Loading and Extracting Data

To successfully load Cifar-10 using tfds.load(), you need to specify two important parameters: as_supervised and batch_size. Here are the steps you should follow:

Step 1: Loading the Datasets

Use the following code to load your datasets:

[[See Video to Reveal this Text or Code Snippet]]

Explanation of Parameters

as_supervised: Set to True to return a tuple structure of (inputs, labels). If set to False, it will default to a dictionary format which is less convenient for CNN practices.

batch_size: When set to -1, it means you want to load all the data at once. This is useful while practicing on smaller datasets like Cifar-10.

Step 2: Extracting Images and Labels

Once you have the datasets loaded, extracting the images and labels can be done in one of two ways: using as_numpy() or by iterating over the dataset.

Option 1: Using as_numpy()

You can simply do:

[[See Video to Reveal this Text or Code Snippet]]

This will convert the dataset into NumPy arrays, which are easy to work with in your projects.

Option 2: Iterating Over the Dataset

If you prefer not to convert to NumPy, you can iterate through the dataset as follows:

If as_supervised is False:

[[See Video to Reveal this Text or Code Snippet]]

If as_supervised is True:

[[See Video to Reveal this Text or Code Snippet]]

Conclusion

Using tfds.load() to access the Cifar-10 dataset doesn’t have to be complicated. By setting the right parameters and following the straightforward extraction options provided, you can easily get your data and start practicing with CNNs. This method allows for cleaner code and better integration with TensorFlow’s data pipeline.

Now you’re ready to take on your CNN projects with the Cifar-10 dataset! Happy coding!