Vision Engine

Applying image recognition in a short timeframe

Are there business processes where image recognition can add value in your company? It is often challenging to answer 'yes' to this, as feasibility can often only be demonstrated after much of the development has been done. This is why Pegamento has set up the Vision Engine. We will walk you through:

Want to know what we can achieve together by deploying the Vision Engine in business processes? Read further or contact us!

What is de Vision Engine?

Lightning-fast custom image recognition software

When it comes to image recognition software, the current market is divided into two segments. Specialised applications are developed for specific niches, and there is the possibility of developing customised applications. At Pegamento, we've seen that the specialist applications are often not easy to apply to non-standard cases, and that customisation carries risk, as a significant investment is required to demonstrate feasibility. Because we develop a lot of customised image recognition software ourselves, we have standardised and fully modularised the most common building blocks in the Vision Engine. Thus, we are able to set up and validate most of a customised application in a short period of time.

A set of standard building blocks

What types of things are already included in the Vision Engine? We distinguish between input, detection & identification, and output modules.

Input modules

As the name suggests, input modules ensure that images can enter the Vision Engine. This could be via a USB or network camera, via an API or web page, or via a folder of videos or photos.

Detection and identification modules

These modules form the core of the software and specifically handle image recognition. They enable images or video frames to be analysed using Artificial Intelligence. But they are also capable of more traditional image recognition. With these modules, we can detect, identify, track, and count things.

Output modules

An application is of little use if the results cannot go anywhere. In the output module, it is possible to write the results to a database, feed them back via an API, or display them on a website or app.

Examples of applications of the Vision Engine

We fully understand that image recognition is a means and never the end in itself. Therefore, we would like to take you through some examples where we have implemented the modules in the Vision Engine:

  • A vision system that inspects the quality of products and rejects automatically
  • A desktop application that analyses images for research
  • A server application that analyses photos and sends them back to an app

The advantages of the Vision Engine

Quickly demonstrate feasibility

We can quickly demonstrate feasibility in a Proof of Concept, because often no development time goes into setting up peripheral issues. Therefore, we see quickly enough whether your application is feasible and what we can expect in terms of accuracy.

Shorter development cycle

Often, we can realise most of an application using standard components that we then only need to configure. As a result, we quickly have the basis of the software in place and we produce a Proof of Concept or Minimum Viable Product in a short period of time.

Higher software quality

Because we use Vision Engine modules frequently, all customers benefit from the ongoing developments we do. This also enables each component to be tested more extensively in multiple situations.

A few client cases

feature matching
feature matching

Assessing the quality of bread for Bakkerij Amstelveld

The challenge for Bakkerij Amstelveld was to be able to recognise when bread rolls are irregular, causing problems in the production line, or not suitable for delivery to customers. Using a USB3 camera as input, an AI network for segmentation, and an AI network for quality control, the Vision Engine ultimately controls a PLC as output to open a reject belt.

Detecting objects on the bridge deck for Rijkswaterstaat

Sometimes unsafe situations occur on the bridge deck. The operators see this and know not to open the bridge deck. With 2 network cameras as input, an AI Anomaly detection network, and a web page as output, the Vision Engine can still assist operators in their decision in case they overlook something.

Measuring people's heights with an app for Ferring Pharmaceuticals

Determining exact height is important for prescribing growth hormones. An app delivers a video of a person to the Vision Engine on a server. Using Structure from Motion technology, the engine builds a 3D model, recognises and measures the person and, as an output, sends the person's height back to the app.

Want to apply the Vision Engine?

The Vision Engine combines the advantages of a standard product with the flexibility of customisation. But to estimate the benefits for a specific case, we always need knowledge of our customers' field of expertise. We are therefore happy to take the time to exchange views on the added value of applications. Contact us for an appointment.

You might also be interested in the following:

SeeingSupports bridge operators in operating bridges safelyRijkswaterstaat (the Dutch Ministry of Infrastructure and Water Management) appointed Pegamento as

Read more
Customer case – Rijkswaterstaat

Artificial Intelligence (AI) is increasingly being used to solve automation issues. Self-learning algorithms are making it easier to

Read more
Is AI the best solution? Back to classic image recognition

Bakkerij Amstelveld is the first bakery in the Netherlands to use a system with a camera (image recognition)

Read more
Bakkerij Amstelveld uses image recognition with AI from Pegamento for bread quality control