imgo 2.5.8

Image Dataset Management Toolkit

imgo

Process, Augment, and Balance Image Data

💡 What is it?

This library is designed to facilitate the preprocessing phase of image classification projects in order to get into the fun part: training the models!

Features:

Imgo is composed of two modules: uptools and augtools.

Uptools helps to streamline various image data preprocessing tasks, such as:

• Reading images from a local disk
• Rescaling images
• Normalizing and standardizing pixel values
• Converting image datasets into numpy-arrays
• One-hot-encoding label data
• Splitting image datasets into training, validation, and testing subsets
• Merging data subsets into a single dataset
• Saving numpy-arrays as images in class subdirectories

Augtools allows the user to quickly and efficiently apply augmentation to image data. With Augtools, users can perform the following augmentation tasks using very few lines of code:

• Apply a powerful collection of transformation and corruption functions
• Augment images saved on a local disk
• Save augmented images in class subdirectories
• Augment entire image datasets
• Augment training data in place in preparation for machine learning projects
• Rebalance class sizes by generating new training images

🛠️ Setup

Install it from PyPI by running pip install imgo.

Dependencies

The code was written with Python 3.6, and it is recommended to run it in a virtual environment. All the required libraries are listed in the requirements.txt file in this repo.

🚀 Execution

Once the package has been installed, it is simply a case of experimenting with the various classes and functions. For a quickstart, please see the demo.

📝 Documentation

Documentation is currently available in the form of docstrings.

⚖️ License

The project is licensed under the MIT license.

Acknowledgements

Some of the augtools library is built as a wrapper around Imgaug, a powerful image augmentation library. For more information, please see https://imgaug.readthedocs.io/en/latest/.