Deep Learning for Machine Vision

Unlocking new horizons with AI

Machine Learning Revolutionizes Quality Control in Automation

Intelligently confronting difficult inspection situations

Artificial Intelligence (AI) methods are proving to be a reliable solution to many problems in optical quality assurance.  Deep learning as a subfield is based on the application of neural networks to make predictions or decisions. In many ways, the way it works is based on learning in the human brain.

Mastering Difficult Challenges

Deep Learning as a tool for complex problems

There is a wide range of tasks in machine vision that people solve intuitively, but which are difficult to describe parametrically. As a result, they cannot be solved by rule-based methods.

This includes the analysis of products with a wide range of features or non-separable characteristics. Deep Learning is also the method of choice for objects with transparent, reflective, curved or inhomogeneous surfaces.

The reason for this impressive capability is the self-learning aspect. A neural network is fed image data and the desired decisions. Using a large amount of this data, the system is able to determine a suitable solution path. The patterns and properties that lead to the decision are derived independently.

It is assumed that the decision path can be considered universal as the statistical variance of the input data increases. This makes it possible to apply the method accurately to the analysis of previously unknown data.

Deep Learning Ensures Product Quality

What kind of tasks does AI solve?

The strength of Deep Learning often lies in application areas where conventional, rule-based methods no longer offer any solution possibilities.

Typical application areas of AI include:

• Detect defects
• Check for completeness
• Read texts or markings
• Sort and count products
• Localize objects
• Control robots

In borderline situations, AI is often the technological solution that makes automated handling possible in the first place - instead of opting for manual inspection procedures for reasons of safety and precision.

AI in the Production Process

We integrate Deep Learning into your automation solution

Over the course of a decade, Deep Learning has evolved from an unimpressive approach in practice to an indispensable technology for demanding optical inspection tasks.

With our experience from numerous industrial projects and our expertise in machine vision, we develop the optimal AI solution for your production process - whether as a turnkey solution or process-accompanying.

When it comes to implementing AI applications, we offer a wide range of services tailored to your needs. We support you throughout the entire planning and implementation phase, right through to integration. Our services range from turnkey test machines, which we design and manufacture, to partial services for planning and integration, to consulting and training.

We are independent in our choice of tools and focus on the optimal technological solution. We use our own highly flexible Deep Learning platform Neuralyze as well as products from other well-known software vendors.

Deep Learning Ensures Product Quality

What kind of tasks does AI solve?

We offer the development of customised AI models for your specific application as a service.

 Whether you are new to this technology and want to build up your expertise with us, or you are an experienced AI user and want to leave the development and training of your models to machine vision experts, you will benefit from our experience of over a decade of successfully implemented application scenarios in industrial image processing and machine learning.

Our service model gives you the freedom to focus on your company's expertise and thus reach your goal faster with your AI application.

 We monitor, maintain and train your model on an ongoing basis, ensuring that your production runs smoothly.

Contact us today.

Deep Learning in 6 Steps

Integrate Vision AI into existing processes

1. In close cooperation with you, we develop the specific Use Case or inspection task on the basis of your production.
2. Image data of the relevant situation is captured and collected. If you do not have the appropriate acquisition hardware, we can provide a suitable acquisition system.
3. As part of a feasibility study, we recommend a model architecture for your task, label the image data and train the model.
4. Based on the successful outcome of the training, you can decide on the project planning of a customized image acquisition system for your use case.
5. Newly acquired data from current production are evaluated with the model of the feasibility study. You decide whether you already use the results or examine them again in more detail as part of a validation period.
6. If the test results prove to be stable even with a large variation in the number and type of parts, there is nothing standing in the way of transferring them to a production system.

Benefits

• Powerful in difficult inspection situations
• Effective on transparent, reflective, curved or inhomogeneous surfaces
• Detects objects even with high feature variance
• Intuitive defect specification via image data instead of descriptive rules
• Clear differentiation of diffuse properties and features
• Comprehensible without expertise in conventional, rule-based image processing
• Large range of possible applications
• Standardized integration capability
• Highly effective, extensively developed technology