The most stand-out development that Bcnvision has presented at the Advanced Factories trade fair is Cognex ViDi.
Cognex ViDi is an artificial vision system based on algorithms that provides machines with the ability to learn by themselves from example. It is based on artificial neuronal networks, attempting to come close to human perception.
It functions based on the training of the system with labelled images that represent the features, anomalies and known classes of the part or product to be analyzed. This supervised training period teaches the system to recognize defects that may come in multiple forms, to locate and classify specific features or objects and to read text or characters. Then, in an unsupervised phase, the system learns the normal appearance of an object, including its significant but tolerable variations. It is a looping process of constant improvement, where the model can be adjusted as many times as necessary and where the result can be validated until it works as desired..
The VisionPRO ViDi software has been developed to analyze real-world industrial images and offers four independent tools (Locate, Analyze, Classify and Read) that can be combined with other Cognex vision tools that enable the location of features or objects, the detection of anomalies and aesthetic defects, the classification of objects and scenes and the reading of challenging text and characters.
Cognex ViDi offers several advantages over Traditional Artificial Vision.
- It is easy to configure, simply train it with images.
- It does not require software development; it adapts quickly to new models without the need to reprogram its main algorithms.
- It tolerates defect variations. It combines the flexibility and the application of logical criteria of human vision with the security and speed of an artificial vision system.
- It resolves complex inspection, classification, and very difficult location applications with classic rule-based algorithms.
If you have a complex application and believe that you could solve it with Deep Learning, please contact us.