Picture Augmentation with Keras Preprocessing Layers and tf.picture


Final Up to date on August 6, 2022

If you work on a machine studying downside associated to pictures, not solely do you should gather some photographs as coaching information, however you additionally have to make use of augmentation to create variations within the picture. It’s very true for extra advanced object recognition issues.

There are lots of methods for picture augmentation. It’s possible you’ll use some exterior libraries or write your individual capabilities for that. There are some modules in TensorFlow and Keras for augmentation too.

On this submit, you’ll uncover how you should use the Keras preprocessing layer in addition to the tf.picture module in TensorFlow for picture augmentation.

After studying this submit, you’ll know:

  • What are the Keras preprocessing layers, and easy methods to use them
  • What are the capabilities supplied by the tf.picture module for picture augmentation
  • use augmentation along with the tf.information dataset

Let’s get began.

Picture augmentation with Keras preprocessing layers and tf.picture.
Photograph by Steven Kamenar. Some rights reserved.


This text is split into 5 sections; they’re:

  • Getting Photographs
  • Visualizing the Photographs
  • Keras Preprocessing Layers
  • Utilizing tf.picture API for Augmentation
  • Utilizing Preprocessing Layers in Neural Networks

Getting Photographs

Earlier than you see how you are able to do augmentation, you should get the pictures. In the end, you want the pictures to be represented as arrays, for instance, in HxWx3 in 8-bit integers for the RGB pixel worth. There are lots of methods to get the pictures. Some might be downloaded as a ZIP file. If you happen to’re utilizing TensorFlow, it’s possible you’ll get some picture datasets from the tensorflow_datasets library.

On this tutorial, you’ll use the citrus leaves photographs, which is a small dataset of lower than 100MB. It may be downloaded from tensorflow_datasets as follows:

Operating this code the primary time will obtain the picture dataset into your laptop with the next output:

The perform above returns the pictures as a tf.information dataset object and the metadata. This can be a classification dataset. You may print the coaching labels with the next:

This prints:

If you happen to run this code once more at a later time, you’ll reuse the downloaded picture. However the different method to load the downloaded photographs right into a tf.information dataset is to make use of the image_dataset_from_directory() perform.

As you may see from the display output above, the dataset is downloaded into the listing ~/tensorflow_datasets. If you happen to take a look at the listing, you see the listing construction as follows:

The directories are the labels, and the pictures are recordsdata saved below their corresponding listing. You may let the perform to learn the listing recursively right into a dataset:

It’s possible you’ll need to set batch_size=None if you do not need the dataset to be batched. Often, you need the dataset to be batched for coaching a neural community mannequin.

Visualizing the Photographs

It is very important visualize the augmentation outcome, so you may confirm the augmentation result’s what we would like it to be. You should use matplotlib for this.

In matplotlib, you have got the imshow() perform to show a picture. Nevertheless, for the picture to be displayed appropriately, the picture must be offered as an array of 8-bit unsigned integers (uint8).

Given that you’ve got a dataset created utilizing image_dataset_from_directory()You may get the primary batch (of 32 photographs) and show a number of of them utilizing imshow(), as follows:

Right here, you see a show of 9 photographs in a grid, labeled with their corresponding classification label, utilizing ds.class_names. The pictures must be transformed to NumPy array in uint8 for show. This code shows a picture like the next:

The entire code from loading the picture to show is as follows:

Observe that for those who’re utilizing tensorflow_datasets to get the picture, the samples are offered as a dictionary as an alternative of a tuple of (picture,label). It’s best to change your code barely to the next:

For the remainder of this submit, assume the dataset is created utilizing image_dataset_from_directory(). It’s possible you’ll have to tweak the code barely in case your dataset is created in another way.

Keras Preprocessing Layers

Keras comes with many neural community layers, resembling convolution layers, that you should prepare. There are additionally layers with no parameters to coach, resembling flatten layers to transform an array like a picture right into a vector.

The preprocessing layers in Keras are particularly designed to make use of within the early phases of a neural community. You should use them for picture preprocessing, resembling to resize or rotate the picture or regulate the brightness and distinction. Whereas the preprocessing layers are purported to be half of a bigger neural community, you too can use them as capabilities. Under is how you should use the resizing layer as a perform to rework some photographs and show them side-by-side with the unique:

The pictures are in 256×256 pixels, and the resizing layer will make them into 256×128 pixels. The output of the above code is as follows:

For the reason that resizing layer is a perform, you may chain them to the dataset itself. For instance,

The dataset ds has samples within the type of (picture, label). Therefore you created a perform that takes in such tuple and preprocesses the picture with the resizing layer. You then assigned this perform as an argument for the map() within the dataset. If you draw a pattern from the brand new dataset created with the map() perform, the picture can be a reworked one.

There are extra preprocessing layers obtainable. Some are demonstrated under.

As you noticed above, you may resize the picture. You can too randomly enlarge or shrink the peak or width of a picture. Equally, you may zoom in or zoom out on a picture. Under is an instance of manipulating the picture dimension in varied methods for a most of 30% enhance or lower:

This code reveals photographs as follows:

Whilst you specified a set dimension in resize, you have got a random quantity of manipulation in different augmentations.

You can too do flipping, rotation, cropping, and geometric translation utilizing preprocessing layers:

This code reveals the next photographs:

And at last, you are able to do augmentations on shade changes as nicely:

This reveals the pictures as follows:

For completeness, under is the code to show the results of varied augmentations:

Lastly, you will need to level out that the majority neural community fashions can work higher if the enter photographs are scaled. Whereas we normally use an 8-bit unsigned integer for the pixel values in a picture (e.g., for show utilizing imshow() as above), a neural community prefers the pixel values to be between 0 and 1 or between -1 and +1. This may be carried out with preprocessing layers too. Under is how one can replace one of many examples above so as to add the scaling layer into the augmentation:

Utilizing tf.picture API for Augmentation

Moreover the preprocessing layer, the tf.picture module additionally offers some capabilities for augmentation. In contrast to the preprocessing layer, these capabilities are supposed for use in a user-defined perform and assigned to a dataset utilizing map() as we noticed above.

The capabilities supplied by the tf.picture are usually not duplicates of the preprocessing layers, though there’s some overlap. Under is an instance of utilizing the tf.picture capabilities to resize and crop photographs:

Under is the output of the above code:

Whereas the show of photographs matches what you may count on from the code, using tf.picture capabilities is sort of completely different from that of the preprocessing layers. Each tf.picture perform is completely different. Due to this fact, you may see the crop_to_bounding_box() perform takes pixel coordinates, however the central_crop() perform assumes a fraction ratio because the argument.

These capabilities are additionally completely different in the way in which randomness is dealt with. A few of these capabilities don’t assume random habits. Due to this fact, the random resize ought to have the precise output dimension generated utilizing a random quantity generator individually earlier than calling the resize perform. Another capabilities, resembling stateless_random_crop(), can do augmentation randomly, however a pair of random seeds within the int32 must be specified explicitly.

To proceed the instance, there are the capabilities for flipping a picture and extracting the Sobel edges:

This reveals the next:

And the next are the capabilities to control the brightness, distinction, and colours:

This code reveals the next:

Under is the entire code to show the entire above:

These augmentation capabilities must be sufficient for many makes use of. However if in case you have some particular concepts on augmentation, you’ll most likely want a greater picture processing library. OpenCV and Pillow are frequent however highly effective libraries that let you remodel photographs higher.

Utilizing Preprocessing Layers in Neural Networks

You used the Keras preprocessing layers as capabilities within the examples above. However they will also be used as layers in a neural community. It’s trivial to make use of. Under is an instance of how one can incorporate a preprocessing layer right into a classification community and prepare it utilizing a dataset:

Operating this code provides the next output:

Within the code above, you created the dataset with cache() and prefetch(). This can be a efficiency method to permit the dataset to organize information asynchronously whereas the neural community is educated. This might be important if the dataset has another augmentation assigned utilizing the map() perform.

You will note some enchancment in accuracy for those who take away the RandomFlip and RandomRotation layers since you make the issue simpler. Nevertheless, as you need the community to foretell nicely on a large variation of picture high quality and properties, utilizing augmentation can assist your ensuing community turn out to be extra highly effective.

Additional Studying

Under is a few documentation from TensorFlow that’s associated to the examples above:


On this submit, you have got seen how you should use the tf.information dataset with picture augmentation capabilities from Keras and TensorFlow.

Particularly, you realized:

  • use the preprocessing layers from Keras, each as a perform and as a part of a neural community
  • create your individual picture augmentation perform and apply it to the dataset utilizing the map() perform
  • use the capabilities supplied by the tf.picture module for picture augmentation

Develop Deep Studying Initiatives with Python!

Deep Learning with Python

 What If You Might Develop A Community in Minutes

…with only a few strains of Python

Uncover how in my new Book:

Deep Studying With Python

It covers end-to-end initiatives on matters like:

Multilayer PerceptronsConvolutional Nets and Recurrent Neural Nets, and extra…

Lastly Convey Deep Studying To

Your Personal Initiatives

Skip the Teachers. Simply Outcomes.

See What’s Inside


Please enter your comment!
Please enter your name here