PyCVF a Computer Vision Framework.
PyCVF is first aim is to be a Framework
It aims at being a framework, meaning a set of conventions and utilities
that makes easy the writing of your application.
Hence its first aim is not being a library or a toolkit.
Although for the framework to be useful a set of libraries and utilities will be provided.
PyCVF is composed of different elements :
- A set of standard datasets that are easily download-able;
- A set of conventions for writing new programs, and extending the framework.
And in a more wide interpretation:
- A library, and a set of wrapper libraries allowing access to essential algorithms in computer vision, statistics, indexing;
- A set of programs aiming at solve classical different tasks : viewing databases, labeling databases, computing features;
PyCVF aims at making computer vision as compact as English
Some framework require you to write hundred lines before to make something interesting.
With PyCVF, we want that what has to be implicit remains implicit in the PyCVf language.
This indeed does not mean that options, and configuration files will not allow the user
to have a very wide variety of behaviors.
PyCVF aims at making computer vision is targeted to scientists
There exists many dataflow based framework dedicated to various fields: PureData,
Max Jitter, Simulink, VVVV, Marsyas and so on... So what makes PyCVF different.
Here is numerous reasons that makes PyCVF different.
- PyCVF is in Python, and thus data format independent.
Most framework are programmed in C, and are thus strongly typed.
- PyCVF allow you to specify and design datasets, it is thus designed for experimental sciences.
- PyCVF supports metadata. PyCVF allow nodes to access metadata, and adapt metadata all along the dataflow.
- PyCVF supports machine learning. Most framework have no real support for machine learning, and the user has many commands to run for tested model that must be trained. PyCVF knows to train, and to organize data required along training.