text.skipToContent text.skipToNavigation

Train AI Models in the Cloud with the AI Lab

Train AI models in the cloud and benefit from higher training stability and scalable performance – regardless of the limitations of local edge systems. With the software and hardware of the uniVision ecosystem, an end-to-end vision solution is created in the AI Loop that combines AI and rule-based methods.

What Tasks Can I Solve with AI Models from the AI Lab?

Quality Inspection

AI-based quality inspection enables accurate detection of deviations, defects and production errors in real time. Even under challenging conditions, products are reliably analyzed and quality deviations are detected early, ensuring consistently high product quality.

Examples: Soldering points, plastic injection molded parts, casting compounds

Object Detection

AI-based object detection enables the precise identification of products, components and objects reliably, even under challenging conditions and with different object variants. With intelligent image analysis, processes can be automated, objects can be clearly assigned and production processes can be made more efficient.

Example: Type detection, sorting tasks, presence check (e.g. of holes, threads and attachments, detection and checking of different types of markings)

Surface Analysis and Structural Testing

AI-powered surface analysis reliably detects scratches, dents, contamination, material defects and other surface deviations in real time. Thanks to intelligent image analysis, the system works precisely even with changing material batches, reflective materials or varying surface structures. This ensures consistently high quality and sustainably reduces waste.

Examples: Scratch analysis, testing of the structure of fabric and non-woven fabric, detection of the rotational position of wooden boards

Foreign Object Recognition

AI-based foreign object control reliably identifies contaminants, debris and unwanted objects. Even with changing light conditions, complex backgrounds or different product shapes, the intelligent image analysis ensures precise and stable detection.

Examples: Empty container inspection, contamination inspection in food

Completeness Check

The AI-powered completeness check reliably detects whether all required components of a product are completely present, correctly positioned and free from damage. Using intelligent image analysis, the system inspects even complex assemblies precisely and in real time. Even in difficult lighting conditions, varying component shapes or demanding backgrounds, the AI ensures stable and safe inspection. As a result, assembly errors, missing parts and production deviations can be detected at an early stage and waste can be sustainably reduced.

Example: Completeness check of bottles in containers

What Are the Benefits of Training AI Models in the Cloud (AI Lab)?

Advantages of Cloud AI Training for Industrial Image Processing

High-performance computing power for high-performance AI models for image classification
Replacement for costly local infrastructure
Flexible access to the AI Lab, allowing projects to be processed anytime, anywhere
Sophisticated user management allows editing of data sets and parts of them simultaneously by different users.

How to Get Started Easily with AI Image Processing?

Intuitive Operation and Seamless Integration into Existing Systems

Seamless connection to wenglor machine vision hardware via the weHub connection tool
Efficient cloud data management lets you start training your AI models right away
Consistent interaction patterns and logical workflows simplify entry without time-consuming training

How to Understand and Optimize AI Models?

Maximum Model Quality through Transparency and Intelligent Analysis

The heatmap function integrated in the AI Lab, the AI input image and score values make AI models comprehensible and understandable
Automatic validation and review function for best possible optimization

How Do I Get AI Models Quickly and Easily?

Seamlessly Connected in the AI Loop. From the First Pixel to the Finished AI Model.

The AI Loop enables easy and fast AI modeling. The AI Lab enables efficient data management and AI model training in the cloud, while the uniVision 3 software enables image recording and the high-performance execution of AI models on machine vision products. The workflow is rounded off by the weHub software as a connection tool and for device discovery.

uniVision 3 in the AI Loop: Capture Image Data and Execute AI Models Locally on Machine Vision devices

As an ecosystem for wenglor machine vision hardware, the uniVision 3 image processing software combines rule-based image processing with AI. It offers a comprehensive toolbox and interfaces.

Direct recording of training data: Image data is captured directly with the machine vision product and uploaded to the AI Lab via weHub
Easy AI model integration: AI models from the AI Lab can be easily imported via weHub and deployed in uniVision 3. The AI models can be directly integrated and executed via the “Image AI” module.
Heatmap display: In addition to the score values per class and the AI model prediction, the heatmap shows which image areas the AI model used in its decision. It is a central tool for better understanding and interpreting the results and evaluating the quality of the AI model.
Numerous standard interfaces: Results can be output directly via standard network-based interfaces or IOs, enabling easy integration into any type of system.

weHub in the AI Loop: Device Discovery and Connection Tool between Local Network and Cloud

The weHub software runs on a separate PC and performs two central tasks: On the one side, it automatically detects wenglor machine vision devices in the local network and makes it possible to carry out all relevant device settings directly; on the other side, it serves as a connection tool between the AI Lab and the local network.

Automatic image upload: Images are uploaded directly to the AI Lab from the uniVision platform for AI model training.
Buffer function: If there is no internet connection, data is stored temporarily.
Deployment: AI models can be downloaded directly from the AI Lab to the uniVision platform.

What Are the Benefits of Interplaying Rule-Based and AI Image Processing?

“The future of image processing lies not in replacing traditional methods with AI, but in combining the two.”
Patrick Schmidt, Product Manager

The demands on industrial image processing are constantly increasing. Growing variety of variants, shorter product life cycles and high quality requirements require solutions that are flexible, powerful and future-proof.

This white paper shows how traditional machine vision and artificial intelligence can be combined in a meaningful way to meet the increasing demands of industrial image processing. Discover how integrated ecosystems support the entire workflow – from data collection to AI training to serial application – and help companies integrate image processing efficiently and scalably into their production processes.

Whitepaper AI in Industrial Image Processing: From Tool to Platform (732 KB)

Learn how traditional image processing and AI can be combined to create a powerful Machine Vision solution.

AI Lab Plans – the Right Plan for Every Requirement!

The AI Lab is used on the basis of prepaid plans. Choose the model that suits you best! The plans differ in terms of duration and scope of resources. The training credits can be used flexibly during the active term. You can book plans consecutively or upgrade within a term as needed. Of course, all plans are set up without automatic renewal – completely transparent and without hidden costs.

	Free plan	Free+ plan	XS plan	S plan	M plan	L plan
Usable storage	1 GB	5 GB	5 GB	15 GB	50 GB	200 GB
Usage term	3 months	6 months	12 months	12 months	12 months	12 months
Training credits	10 credits	20 credits	40 credits	150 credits	500 credits	2,000 credits
Costs	Free of charge	Free of charge*	Subject to charge	Subject to charge	Subject to charge	Subject to charge

*Once in conjunction with the registration of a device with the uniVision AI license package

Free plan
Usable storage 1 GB	Usage term 3 months	Training credits 10 credits	Costs Free of charge
Free+ plan
Usable storage 5 GB	Usage term 6 months	Training credits 20 credits	Costs Free of charge*
XS plan
Usable storage 5 GB	Usage term 12 months	Training credits 40 credits	Costs Subject to charge
S plan
Usable storage 15 GB	Usage term 12 months	Training credits 150 credits	Costs Subject to charge
M plan
Usable storage 50 GB	Usage term 12 months	Training credits 500 credits	Costs Subject to charge
L plan
Usable storage 200 GB	Usage term 12 months	Training credits 2,000 credits	Costs Subject to charge

*Once in conjunction with the registration of a device with the uniVision AI license package

uniVision AI – License Packages

For the execution of the AI models created in the AI Lab, the “Image AI” module is available on the relevant machine vision hardware. This is included in the uniVision AI license package.

The uniVison AI license package includes the activation of the following modules:

“Image AI” module (for AI models from the AI Lab)
“Image ONNX” module (for AI models in ONNX format)

The uniVision AI license is available as follows:

For B60 series smart cameras: DNNL031 license
For MVC series machine vision controllers: DNNL032 license
Device variants of the B60A or MVCA series include the license as standard

Because Trust Builds on Security – We Protect Your Data!

The AI Lab only uses European servers. Automatic backups in the cloud ensure that your data is backed up regularly. Create flexible access restrictions for your images.
Opt for a paid plan and stay in full control of your data!

Two Ways, One Goal: Run AI Models on wenglor Hardware

With the uniVision AI license package, AI models can be run both directly from the AI Lab and in ONNX format directly on wenglor devices – for maximum flexibility in AI model development.
With uniVision AI, your AI models are seamlessly integrated into uniVision. Training can be delivered either end-to-end in the AI Lab or in your own toolchain using common machine learning frameworks such as PyTorch or TensorFlow.

AI Model Training in the AI Lab: “Image AI” Module

Ideal for those who want to leverage the entire AI workflow from a single source.

Model training	Cloud-based in the AI Lab
Data flow	Transfer of images from uniVision to the AI Lab via weHub
Transparency	Heatmaps and evaluations for maximum traceability
Access	Intuitive user interface – ideal for structured projects without their own machine learning toolchain

Externally Trained AI Models in ONNX Format: Module “Image ONNX”

Perfect for anyone already training with open-source frameworks AI models.

Model training	Local or external in own toolchain (e.g. PyTorch, Tensorflow)
Interface	Import of AI models in ONNX format (via Github)
Transparency	Support for heatmaps (depending on the respective AI model)
Integration	Direct execution in uniVision – compatibility and conversion according to documentation

Powerful Hardware for Optimal Execution of AI Models

The AI models trained in the AI Lab are perfectly matched to wenglor hardware. On the B60 series smart cameras, the adaptation of the AI models to the Neural Processing Unit (NPU) ensures their efficient and fast execution. The multicore processor and Intel® OpenVINO™ acceleration of the MVC series machine vision controllers are ideal for running multiple AI models in parallel as well as using different process instances in complex applications.

Smart Cameras B60

Find out more

Machine Vision Cameras with Machine Vision Controller