Deep Learning
ARTIFICIAL INTELLIGENCE FOR COMPLEX INSPECTIONS
Traditional 2D vision has limits in the applications that can be programmed. Deep learning makes it easy to overcome this problem by creating a solution based on learning. The goal is to show examples (samples) of what we want to characterize and the VIDI software transforms these examples into a robust vision application. It's a bit like showing a child several pictures of a car so that he can recognize them by himself in the street after having seen a few examples.
WHEN TO USE DEEP LEARNING?
Deep learning is used to replace operations where the human eye excels: subjective or complex applications. Deep learning is not capable of measurement (fortunately, traditional vision is there for that) but it is perfectly adapted to control aspect, object recognition and classification of defects.
ASPECT CONTROL
Analysis of aesthetic defects of parts in comparison with some samples parts. The system can make the difference between an expected texture (fabric or fibers for example) and a defect in this texture.
COMPLEX PROCESS CONTROL
Analysis of the quality of an operation from a database.
Adaptable for all types of operations: welding, brazing, joint casting... These applications are often complex to program because of the natural variations in the processes.
CHARACTER READING
Reading additional information marked on parts. Here the interest is to identify and assemble strings of characters and record information in order to identify a product or to record its technical information.