realvirtual.io AIBuilder
Develop, train, test, and deploy your AI vision-based industrial process.
Last updated
Develop, train, test, and deploy your AI vision-based industrial process.
Last updated
Training AI models in a digital twin environment offers significant benefits by replicating real-world systems virtually. This approach enables safe, efficient, and scalable AI development.
Key advantages include:
Cost Efficiency: Reduces expenses by avoiding the need for physical setups and minimizing downtime.
Safety: Eliminates risks associated with real-world testing, especially in hazardous applications.
Rapid Data Generation: Synthetic data can be generated in minutes, allowing AI models to be trained on diverse scenarios quickly—whereas collecting equivalent real-world data could take months.
Accelerated Development: Allows rapid iteration and testing, speeding up the AI development cycle.
Seamless Integration: Facilitates real-time testing with automation systems like PLCs, robots, and sensors.
Scalability: Easily simulates complex or large-scale systems for diverse AI training scenarios.
The realvirtual.io AI Builder is a platform designed to simplify and accelerate the development, training, testing, and deployment of AI-driven solutions for industrial processes. Utilizing Unity’s real-time 3D engine and cutting-edge AI technologies like deep learning and convolutional neural networks, AI Builder enables the creation of synthetic training data, advanced simulations, and seamless integration into automation systems. This document will guide you through each step of leveraging AI Builder to optimize industrial vision-based applications.
The AI Builder workflow shown in the image outlines the process of creating and deploying AI-based solutions using realvirtual.io AIBuilder and Unity.
Here is a breakdown of each step in the workflow:
CAD:
This step involves designing the product or machine in CAD software. The 3D data generated from this design is used as the foundation for creating synthetic training data.
Generate Training Data:
Once the CAD data is available, the AI Builder automatically generates 3D, photorealistic, randomized, and annotated training data.
This process is fast, allowing the generation of around 1000 pictures in just 2 minutes.
The training data is saved with associated annotations (such as in .bmp or YAML format), and it is ready to be used for training the AI model.
Train:
With one click, the AI model can be trained using the generated training data within the realvirtual.io environment.
The trained AI model is then exported as a standard ONNX model, which is a popular format for machine learning models that allows for interoperability between different platforms.
Test:
After training, the AI model can be tested within the realvirtual.io Professional platform.
The AI model is tested in a full Digital Twin setup, which includes automation interfaces, PLC systems, robotics, and other components. This ensures that the AI model works as expected in a virtual environment before being deployed to a real-world scenario.
Deploy:
The trained neural network is then deployed as an executable application that runs on platforms like Windows, Linux, macOS, and iOS or it can be deployed as an ONNX model to a system which is able to run the ONNX model.
This deployment can also include industrial communication interfaces such as S7 TCPIP, OPCUA, ADS, Websocket, MQTT, and more.
AI Builder from realvirtual.io is built on Unity 2023.2 and above, requiring a valid Unity license and installation on Windows systems. It delivers a fully prepared project, including a complete AI training and testing pipeline based on Convolutional Neural Networks (CNNs) and PyTorch.
With just one click, you can run the entire AI process seamlessly. Upon purchasing AI Builder, you will receive detailed instructions on accessing the project and utilizing it for your industrial AI solutions.
Besides the downloaded realvirtual.io AI Builder project and a Unity license no additional installation is needed.
