PyYel 0.0.10

New build

PyYel

PyYel is a personnal library that aims at helping the deployement of strong data science tools, from data handling to deep learning.

To-do: (order of priority)

• Implement image augmentation
• Implement a support of pretrained neural networks
• Standardize the use of config files
• Deploy GUI (Tkinter) utilities to pilot the library
• Deploy HTML (dagster) utilities to pilot the pipelines
• Add TensorFlow support

Continuous development :

• Add new models of neural networks
• Improve the support of other data types

Data

This module of the library is dedicated to data management. It possesses powerful utilities to manipulate the data, convert it, augment its size... It is also a good starting point to start implementing a data pipeline that will be fed to a neural network model.

Datapoint

A collection of classes that allow to convert a datapoint to a standardized format, that is expected by all the modules of the PyYel package.

• Datapoint : a class of usefool tools that simplifies the management and handling of a dataset. The inputs are converted to formats usable by the rest of the library.
• Datatensor : a class of useful tools that simplifies the data preprocessing steps, in order to then deploy a machine learning solution using it.

Utils

A collection of powerful tools that permit an easy manipulation of the datapoints.

Augmentations

A compilation of classes featuring methods to augment a datapoint of various type.

• ImageAugmentation : features a handfull of functions that can augment any type of data, as well as its labels. Can be deployed using the dedicated pipeline controller from the Utils.py file.

Networks

Networks regroups all the deep learning aspects of the PyYel package. It defines powerful models and pipelines that allow to easily deploy a machine learning tool to tackle a wide array of tasks and datasets.

Compiler

• Trainer : A training loop wrapper, that performs most of the steps required to train a model.
• Tester : A testing loop wrapper, that performs most of the steps required to test a model.
• Loader : A model loading utility, that allows to easily retreive a saved model from its weight, and to deploy it somewhere else.

Models

• CNNx2 : a simple two-layers convolutional network, that aims at solving multi-labels classification tasks
• CNNx3 : a simple three-layers convolutional network, that aims at solving multi-labels classification tasks
• ConnectedNNx3 : A simple 3 layers fully connected neural network, that can handle a wide range of tasks.
• ConnectedNNx5 : A simple 5 layers fully connected neural network, that can handle a wide range of tasks.