How it works

Deepyt Core

• Python programmed software (3.8.10)
• Multiplatform (Windows 10/11, MacOS, Unix, edge devices (?) )
• Full compatibility with Tensorflow, Pytorch, Sklearn and other opensource artificial intelligence libraries.
• Implementation, execution, saving of acyclic direct graphs (DAG). A graph is represented by an undefined sequence of operations encoded in Python code.
• Optimised graph execution: only the necessary parts are executed
• Modular postprocessing: possibility of developing many calculation and visualisation solutions with dedicated tools. Possibility of customising the dashboard.
• Project-based structure: the user works within a path, containing some mandatory project files and his working documents.
• The software has a node-and-edge based structure, typical of dataflow programming
• A node can represent any programming object, from simple classes or functions to complex operations, or another node realised with Deepyt.
• A graph can be realised via the graphical interface or via the Python library
• For the execution of a graph, the DeepytCore library is sufficient.
• It is possible to join several graphs together, create concatenated graphs, nested graphs, or execute only parts of them, as required.
• The module (node) based structure allows the programming flow to be managed from very low-level operations, moving on to higher level operations. Each correctly programmed module will be reusable.
• The structure of the graph provides a series of inputs (placeholders), operations and Outputs. Only the specified outputs are calculated.

Built-ins, libraries, custom code

• At the lowest level of use, Deepyt can be used to create work chains in Python code.
• It is possible to use all built-ins of the Python version that are compatible with the released version of the software.
• Deeplabs provides several ready-touse libraries, wrapped in graphical versions (numpy, Tensorflow, PyTorch…), with a potentially growing catalogue.
• A user can create an indefinite number of libraries (in the form of custom nodes) which he can choose to import or not into the project folder.
• Libraries can be downloaded or uninstalled at any time, you just need an internet connection and a valid licence.
• The user can use automatic node wrap to transform his classes and functions (or entire python modules) into graphical nodes that can be used directly in Deepyt.

Model development

• Develop AI models without a line of Code, with a node-based structure
• Mixed No-Code / Code approach for advanced models’ implementation
• Development, training, evaluation, testing and results visualization

Growing Gallery of Complete Projects

• Gallery of ready-to-go examples for AI and Engineering applications
• Easily apply examples to your case study changing few parameters

Handling servers

• Run and handle multiple Jupiter servers
• Develop with your Notebooks and Python code, using all the pre-implemented libraries and DeepytCore

Dashboards Builder

• Create custom Dashboards with a growing gallery of available tools for data visualization, analysis and processing
• Every tool works as a standalone Qt Based software: it’s possible to cover a wide range of applications