In addition to the precision and scalability typical of classical systems, it combines the innate human capacity to analyze variations, thanks to image analysis.

A project of this type needs a series of specific phases, in order to properly deal with both the «training phase» and the following «implementation phase».

It is characterized by three phases:

1. As with all applications, in the vision field there is a first phase where preliminary information useful for addressing the application is collected.
2. To make sure that the Deep Learning algorithm responds to the application needs, it is essential to carry out a proof of concept which is then carried out in the “training” phase: in this phase the images are collected and divided between those with good pieces and those with defective pieces. The defects are indicated and highlighted for classification.
3. Implementation phase: acceptance of the technical solution; installation of hardware and production of images, and finally programming of the system.

The images must be made, labelled, the parameters of the algorithms set, a first training carried out, the results validated and possibly refined.

This approach allows the knowledge of vision specialists to be integrated with the experience of those who live production every day. This combination of skills ensures a much more flexible and highly effective level of control.

Contact our team specialized in the world of vision

Edge Learning

Edge learning is a subset of artificial intelligence (AI) where processing takes place on the device, or “at the edge” of the data source, using a set of pre-trained algorithms. The technology is simple to set up, takes less time and requires fewer images for training than other AI-based solutions such as deep learning.

Edge learning is a perfect solution for those who have to manage an application that has too complex variables for traditional vision systems but at the same time does not need to use a deep learning solution.

How does it work?

The first step in designing an edge learning system is done by Cognex, which develops the technology and integrates it into the smart camera to make it available to customers, which complete the training process by adding useful images for their specific application. This process guarantees more speed and results even with few images, also does not require a computer with a GPU, standard computers are sufficient without the GPU.

The benefits of Edge Learning

1. It is much less expensive to implement than rules-based machine vision and deep learning solutions.
2. It allows for faster production and recipe changes because training and production take place in the same place, on the same device.
3. Those using edge learning technology need fewer images and less time than deep learning to train the camera.
4. Edge learning simplifies rule-based machine vision applications because it does not require a specialization in algorithm knowledge.
5. Edge learning is more user-friendly than traditional solutions, the modern platforms In-Sight Vision Suite allow you to easily manage the application.

Contact our team specialized in the world of vision

L'articolo Deep Learning & Edge Learning proviene da Azzali Elettronica.