camera

Computer Vision

How often do you make decisions based on what you can see? Extremely often. This is a frequent occurrence during production processes, in business processes and even as part of our general lives. Automating such decisions can lead to huge efficiency gains.

Within the field of Artificial Intelligence (AI), converting numbers, sounds, text and images into something that can be understood by computers allows us to draw conclusions or take action on that basis. We will take a closer look at static and dynamic images. A camera views an item, reads the image and links this to an intelligent, learned response automatically. This is Computer Vision, the technology of understanding what can be seen in an image.

Lothar Verledens

The latest developments in Computer Vision have finally allowed us to automate a vast range of tasks that rely on the simple act of viewing something to make decisions.

Lothar Verledens | Computer Vison Expert @ Inetum-Realdolmen

Added value in practice

production

Quality control

Inspect production processes. Determine whether products contain defects based on various visual characteristics.

behavior

Behavior monitoring

Identify groups of people such as hikers or cyclists to be able to monitor their behavior. Other examples include fall detection in hospital rooms or compliance with social distancing in shopping areas.

documents

Reading documents

Check for the presence of certain elements in paper documents and make them machine-readable.

How does it work?

To a computer, an image is nothing more than rows of numbers representing specific colors or levels of intensity.  The concept of shapes, as used by people in interpreting images, is difficult for a computer to replicate.

'Preprocessing' and 'feature engineering' help give the computer an idea of what is there to be seen. Images are analyzed and transformed in various ways, and their intrinsic features are extracted. These intrinsic features more closely resemble how a person sees images, and can provide the computer with input relating to shapes and/or textures.

AI can use such features to build an understanding of what is going on in an image, step by step and based on examples.

The ultimate result is a usable model to which we can, for example, present new images for interpretation via cameras. 

Our role as your trusted partner

There are many previously-trained public models available. However, professional use often requires further customization. We tend to use the public models as a starting point, adapting them to our customers' needs. This allows us to make use of the latest, state-of-the-art technology and reap the benefits of transfer learning. An additional advantage of this method is that relatively few images suffice to retrain a model to fit a specific use case. This way we can achieve good results in only a limited time. We have opted for efficient, fast and affordable customization.

We can support you throughout your journey. From the inspirational stage, when we investigate opportunities to create value from your data for your organization using AI together, to co-creation of a first minimum viable product (MVP).

Sign up for our newsletter

Would you like to receive our newsletter and stay informed about your preferred topics? 

Sign up here