Machine vision systems with artificial intelligence
Classical machine vision was and is superior to human capabilities in terms of speed, repeatability and accuracy and achieves best results in quantitative measurement of structured scenes.
In contrast, humans are suited to learning by observing examples or specifications of test parts. In this way, he can distinguish between slight errors in appearance, such as those caused by perspective distortions or differences in brightness, and actual functional errors. It can assess for itself whether a change in appearance affects the required quality.
AI technology uses neural networks to achieve learned knowledge that enables it to distinguish between anomalies, shapes and characters, while tolerating natural deviations. Thus, artificial intelligence combines the superior flexibility of humans with the performance of a machine system.
How do I get a fully trained AI model?
Images are required to train the AI. These are usually real camera images that are collected for this purpose. Alternatively, synthetic images can also be generated from simulations.
Information must be added to the image data, e.g. positions and areas of relevant image information, as well as the type of objects in the image, in the form of classes. Various platforms are available for this: Online via VisionCockpit, offline with VisionTools V60 or when generating images in the simulation.
The model is trained with this data set. VisionCockpit provides the user with various pre-trained neural networks that can be used to train an optimally functioning model for the required evaluation. When using our VisionCockpit cloud service, an evaluation in the form of model testing is also possible.
How do I analyse images with AI methods?
The trained AI model is transferred to the VisionTools AI Box evaluation unit with integrated web server.
The AI box communicates with the peripherals via a REST interface. Requests, including the image, are transmitted via http protocol and the AI evaluation results are available a few tenths of a second later. Several models can be stored on the AI Box. Which model is used can be controlled with the enquiry command.
The VisionTools V60 evaluation software already contains ready-made, configurable objects for communication with the AI Box and is ideally suited for image acquisition, controlling the AI Box, handling communication with the PLC, visualisation and image storage according to various criteria.
AI tools by VisionTools
For the use of artificial intelligence in machine vision projects, VisionTools offers ALL tools for the installation of a Deep Learning inspection system.
Image acquisition
Labelling
AI Training
AI Evaluation
Further information can be found here in our current product folder (PDF, 1033 KB).
Upgrade your system
Do you have questions about solutions in the field of artificial intelligence?
We will implement a system based on AI for you or help you to successfully use deep learning technologies yourself in your application.
To train your data, our cloud service VisionCockpit is available, which allows you to start immediately without the need for any expensive hardware of your own.
Contact us – we will be looking forward to inform you!
+49 (0) 7254 9351 400
Cloud Service by VisionTools – AI Training / File management
- AI training on a server hosted in Germany without own expensive hardware.
- The training is carried out with high-performance GPUs (NVIDIA T4) and is scalable as required, i.e. with VisionCockpit several trainings can be carried out simultaneously.
- The user has various pre-trained algorithms for object recognition at his disposal, with which VisionCockpit can train an optimally functioning model for any application.
- Upload any amount of labelled image data. The training data is stored on the server.
- The Data security of VisionCockpit is TISAX certified.
- Simple, fair billing system.
- The VisionCockpit dashboard will expand to include other useful AI tools in the future.
Multicore GPU Workstation for AI-based Evaluations
- Integrated web server for configuring and uploading the AI training models.
- The image processing application communicates with the AI Box via the REST interface
- The design as a separate module makes it possible to upgrade even older systems with artificial intelligence
- Edge device in robust aluminium housing (IP67) with 2 x Gigabit Ethernet port, power supply 24VDC or PoE, industrial-grade M12 circular connector