Omnis 0.0.10.15

Deep Learning for everyone

Omnis

Deep Learning for Everyone

---

You have just found Omnis.

Omnis is a library of deep neural network applications, written in Python and capable of running on top of Keras and Tensorflow. It was developed with a focus on enabling fast application of deep learning.

Use Omnis if you need a deep learning library that:

• Is EASY to learn.
• Allows for easy and fast use.
• Supports CNN, LSTM, GAN applications.

Omnis is compatible with: Python 3.6-3.7.

---

Deep Block

Omnis has been developed as a backend library of Deep Block. Deep Block is a platform where anyone can use AI technologies with ease. try Deep Block.

---

Getting started: Implement a deep learning application with 4 lines of code!

The core data structure of Omnis is Application which is designed to be easy to use in each field.

Here is an Image Classification example with the Caltech 101 dataset:

from omnis.application.image_processing.image_classification.image_classification import Image_Classification

Choose an application:

image_classifier = Image_Classification(model_type = 'densenet121')

Prepare data:

image_classifier.prepare_train_data(get_image_from='directory', data_path='101_ObjectCategories')

After preparing data, you can train your application:

image_classifier.train(epochs = 40, batch_size = 16)

Now you can use the application to classify images:

prediction_result = image_classifier.predict(data_path = '101_ObjectCategories/accordion')

print('predict labels')
print(prediction_result)

For a more in-depth tutorial about Omnis, you can check out:

In the examples folder of the repository, you will find more applications.

---

Installation

Before installing Omnis, please prepare NVIDIA GPU(s) and install TensorFlow GPU and Keras using conda.

Then, you can install Omnis itself. There are two ways to install Omnis:

• Install Omnis from PyPI (recommended):

If you don't use a conda virtual environment, you can run the command below (not recommended):

sudo pip install omnis

If you are using a conda virtual environment, you may want to avoid using sudo:

pip install omnis

• Alternatively: install Omnis from the GitHub source:

First, clone Omnis using git:

git clone https://github.com/mkh48v/omnis.git

Then, cd to the Omnis folder and run the install command:

cd omnis
python setup.py install

---

Guiding principles

• Simplicity. Omnis pursues a simple architecture. Designing a software with simple architecture not only helps you to understand the code easily but also helps your painful debugging.

• Easiness. Don't worry about complicated algorithms or theories or mathematics. Omnis will handle difficult stuffs for you. Just learn how to use deep neural networks and USE THIS!

• Modularity. No spaghetti code!

---

Why this name, Omnis?

Omnis means EVERY in Latin. The goal of Omnis is to make deep learning technologies easier so that EVERY one can use deep learning technologies without headache.

---