We perform feasibility studies for our customers on daily basis. We use them to demonstrate the capabilities of the PEKAT VISION software. On two examples of such studies we want to show you where and how exactly can **PEKAT VISION's focused learning** algorithm be applied.

This time we aimed at the **food industry**. Namely, we chose puffed rice bread and biscuits to show you two possible approaches to quality control and defect detection using PEKAT VISION software.

The first example shows a very similar task on the control of biscuits using the **Surface module**.

![Surface module](/media/8/image4.png)

Working with the Detector module in our user interface is very intuitive. When using this module, we literally trace the relevant characteristics (typically defects, but also others) that we want to detect. The characteristics of the examined objects can be divided into a number of classification classes.

After the annotation, it is necessary to train the algorithm to detect the relevant characteristic.

The result might look like this:

![Surface module result](/media/8/image2.jpg)
*(The picture shows the result of detection when using one classification class - "Defect"; “Vada” in czech.)*

The second example shows the use of the **Detector module** to check the integrity of puffed rice bread:

![Detector module](/media/8/image3.png)

The goal is to mark objects and divide them into classes. In the described case, we divided the puffed rice breads into two classes - good (OK) and bad (BAD). However, an unlimited number of classification classes can be defined, just like when using the Surface module

After classification, it is necessary to teach the algorithm to detect the relevant characteristic of the class in multiple images showing the same type of characteristic (typically a defect, but also others like colour).

The result of the detection can then look like this:

![Detector module result](/media/8/image1.png)
*(The picture shows the result of detection using two classification classes - "OK" and "BAD".)*

The examples show that multiple approaches can be chosen for similar tasks. To assure that the users of the PEKAT VISION software do not have to waste their time with searching for the right solution, we offer the possibility of conducting a Feasibility study on the submitted images, which you send to us directly (you can use directly this [link](https://pekatvision.atlassian.net/servicedesk/customer/portal/4/group/5/create/28) or to our partners and international distributors (you can see the list of them on the bottom of this site: [https://www.pekatvision.com/contact](https://www.pekatvision.com/contact)).

You can find more detailed explanations of functions and video tutorials on our website [https://www.pekatvision.com/videos](https://www.pekatvision.com/videos) or on our youtube channel [https://www.youtube.com/channel/UCvbx26TqHNgbIVAY-80XJ1A](https://www.youtube.com/channel/UCvbx26TqHNgbIVAY-80XJ1A).



pills.jpg

We perform feasibility studies for our customers on daily basis. We use them to demonstrate the capabilities of the PEKAT VISION software. On two examples of such studies we want to show you where and how exactly can PEKAT VISION's focused learning algorithm be applied.

Case study: biscuits and puffed rice bread

Hirokazu TERADAIRA
Senior General Manager / Sales / ITANZI / TIETECH Co. LTD. / Japan

  **Could you introduce TIETECH? (activities, strategy, main industries, future plans)**

Our company provides products tailored to various machine tools by making full use of
motor controls and controller-related technologies cultivated over many years. Based on
electronic application technologies such as microcomputers, we develop solutions related to
controllers, imaging devices and IoT (Internet of Things) solutions.
We also manufacture products that meet our customers’ needs. We specialize in FA
(Factory Automation) products, provide optimal solutions, and widely contribute to society.
We offer solutions not only for visualization of equipment and production, but also for on-site
improvement and management decision support, which aggregate and analyze information
about people, things and machines.
We provide products that utilize camera images as &quot;eyes&quot; by making full use of image
processing technology and high-speed, large-capacity communication technology.

**TIETECH and PEKAT VISION have signed distribution contracts for Japan. Why did you
decide to start distributing PEKAT VISION deep learning software?**

Because we thought that PEKAT VISION would suit the FA industry field, which is our main
market in Japan.

**In which industries do you see the biggest potential for the PEKAT VISION deep learning
software?**

We see a lot of potential in the FA industry.

**How would you describe the deep learning market in Japan?**

This is a new market in Japan, but it is up-and-coming. There are many opportunities
arising day by day, and as a leading company we need to be at the front of this.

**How do you see the future of the Japanese deep learning market?**

We now need solutions to solve issues such as labor shortages and craftsmanship
succession problems in FA industry field in Japan. So it is the right time to introduce
useful AI software such as PEKAT VISION.

**What interested you the most about PEKAT VISION? What do you think will be the biggest
benefit for your customers?**

We are interested in the following features of PEKAT VISION:

- Various inspection modes and functions.
- Less training data required.
- Fast evaluation and high accuracy.
- Possibility to be embedded with our NVIDIA Jetson TX2 and Nano carrier boards

We expect that it will be suitable for complex inspections in the FA industry, as well as for
small-quantity &amp; high-mix product inspection.


/media/7.png

TIETECH - new distributor in Japan

PEKAT VISION automated inspection process of brake discs production in Accenture Garching IIoT Innovation Center. Using high-end anomaly detection module we were able to automate the process of brake discs inspection.

/media/4/small_media.jpg

johan_kastensson_preview.jpg

PEKAT VISION Brake Discs Inspection at Accenture with Deep Learning



Detecting hair in jelly using Pekat Vision's Detector module trained with annotated images
Czech firm Pekat Vision writes about quality control projects it has solved in pharmaceutical and food inspection, both key markets that need to be kept running during the Covid-19 pandemic

Fighting crisis with machine vision

PEKAT VISION  and OEM Automatic AB established a partnership in 2019. We set down to talk with Sales Manager of OEM Johan Kastensson to understand what drove this decision and how it’s going to impact the machine vision market. 

**What is your strategy for accepting new suppliers?**

The goal for OEM is to add value to our customers. This means we need to offer the top of the line hardware and software, to provide the best solutions for our customers, to be their trusted consultants. We need to have the right tools to assist with a very diverse range of sophisticated customer projects. 

**Your company decided to add PEKAT VISION software to its portfolio. Why did you choose PEKAT VISION?  What led you to this decision?**

As we were doing market research for top tear machine vision software we found some information about PEKAT vision on the web. Coincidentally, we heard positive feedback about PEKAT from our partners in the field around the same time. So we decided to reach out directly to check out the software; and possibly do a proof of concept.  We saw the company had a positive profile and the people behind it were quite capable. When all testing was completed successfully, the decision was made that this is something we should provide to our customers in Scandinavia.

**What interested you the most about PEKAT VISION, what do you think will help your customers the most?**

One of the key features which I find the most interesting, is the ability to run a proof of concept very quickly. This means that we can show a customer if a project is feasible or not even at early inception stages. Very different than what has been done in the past with the traditional approach. 

Our staff at OEM is very knowledgeable on cameras and images.  We are able to leverage our years of expertise in the industry to help customers get the best images in order to get optimal performance out of the PEKAT VISION software.  

Since the software is capable of running in Windows environment, it is really simple to deploy it in a familiar environment. We also enjoy the flexibility of the software to add additional features and logic using our own code.

**What do you see as the biggest challenge for companies starting with PEKAT VISION and what is your advice for them, how to deal with it?**

The main obstacle is dealing with something new. Not everyone is familiar with this type of technology. It takes time, and sometimes more than the customers initially expect.  However, OEM is here to help the customers as the first line of assistance. We know the software and can guide them through it. In case there is a more complicated task - we have great support from PEKAT VISION.  Everything starts with a proof of concept. PoC is our assurance that the solution will work for their project for both testing and future deployments. 

**Last but not least, what is your opinion on the increasing demand for deep learning solutions on the current market?**

I think deep learning is taking over the industrial world. With volumes needed to cover global demands, now more than ever, companies need to be cost-effective and find ways to save money within production processes. Software like this can help customers to develop sustainable and scalable solutions for the future. Just watch. It’s not hype, it’s a revolution. 


PEKAT VISION  and OEM Automatic AB established a partnership in 2019. We set down to talk with Sales Manager of OEM Johan Kastensson to understand what drove this decision and how it’s going to impact the machine vision market. 


Media

Case study: biscuits and puffed rice bread

TIETECH - new distributor in Japan

PEKAT VISION Brake Discs Inspection at Accenture with Deep Learning

Fighting crisis with machine vision

Interview