News Detail

Reprint【Three Companies Form Business Alliance】Sirius Vision, Marushin, and Tactpixel Strengthen Joint Development Structure to Commercialize "mikaesu," an AI Label Calibration Tool to Promote DX in the Food Industry

(hereinafter referred to as "Marushin") and Tact Pixel Corporation (hereinafter referred to as "Tact Pixel") have formed a business alliance for the purpose of promoting DX (Digital Transformation) in the food industry.

Through this partnership, Marushin will integrate its practical know-how in food package manufacturing with the AI calibration tool "mikaesu," currently available free of charge on food-labeling.com, and begin full-scale discussions and joint development toward future commercialization and sales.

 

Background and purpose of the alliance: To eliminate the "complexity" and "gentrification" of food labeling checks.

With the recent revision of the Food Labeling Law, stricter allergy labeling, and diversification of consumer needs, the contents of food labels are becoming more complex every year. On the other hand, label checking (proofreading) at manufacturing and design production sites still relies heavily on the visual inspection of skilled personnel, increasing the workload and risk of human error.

In response to these issues, SiriusVision, with its strength in image inspection technology, Tact Pixel, an AI technology developer, and Marushin, with its knowledge of food labeling on the reverse side of food packages, have joined forces to create a solid development system to make the AI calibration tool "mikaesu" a "professional product" that can meet the strict standards of actual business operations. The three companies have built a solid development system to transform the AI calibration tool "mikaesu" into a "product for professionals" that can withstand the rigorous standards of actual business operations.

This collaboration will not be limited to the mere provision of technology, but will enable the realization of a "practical AI calibration system" that is not just a theory on paper, but a fusion of knowledge from the field.

 

About the AI proofreading tool "mikaesu

mikaesu" is a cloud tool that reads image data of food labels and labels and supports AI to check the consistency of descriptions and laws and regulations. It is currently available free of charge as a beta version on the web, and we have received feedback from many users.
Through this partnership, Marushin's knowledge of food labeling (optimal solutions for check flow and perspectives of on-site personnel) will be reflected in the development roadmap, with the aim of completing a more robust package for business use.

<mikaesu free public version URL
*Only sake, shochu and wine

Food labeling AI automatic proofreading check "mikaesu

 

Roles of each company in this alliance

By combining the strengths of the three companies, we will provide a full range of services from development to sales and demonstration testing.

 

 

Adopted technology: AI calibration engine "Koallect

The mikaesu engine is powered by Tact Pixel's proprietary Koallect core engine, an AI calibration engine that automatically reads the contents of all package labels, including food labels, and checks them against laws, fair competition codes, and industry voluntary standards. The AI calibration engine "Koallect" is a calibration support system that automatically reads the contents of all package labels, including food labels, and checks them against laws, fair competition codes, and industry voluntary standards. There are many elements on packages that tend to degrade the accuracy of general OCR, such as small capitals, close alignment, low contrast between vertical and horizontal, and low contrast between the package and background. This product combines multiple OCR engines optimized for each application to achieve highly accurate reading that prevents misreading at the single character level. After extraction, the text is structured into semantic units based on the rules for organizing dictionaries, required items, and expression fluctuation, so that even packages with inconsistent formats can reliably extract the information necessary for proofreading. The final output is a report format that organizes the extracted image of the relevant section, the read content, the collation result, and the supporting text, and is provided in a form that is easy for people to make final judgments. The internal engine is implemented in Python, and another key feature of the system is its architecture that allows flexible integration with external UI and business systems.

 

Future Development of the Alliance

In the future, we will continue to enhance the functions under the supervision of Marushin, and together with the knowledge gained from the free public version, we aim to commercialize "mikaesu" (paid packaged version). Sirius Vision will take the lead in the project, and will discuss specific sales schemes and product specifications to contribute to the safety and security of the food industry as a whole, as well as to operational efficiency.

 

Representative Comment

Junichi Tsujitani, Representative Director, Sirius Vision Inc.
As an imaging professional, I am pleased to be taking on the longstanding challenge of DXing the food industry's calibration operations with a strong partner."

Yoji Hiraki, Representative Director, Marushin Co.
We see great potential in this collaboration, which combines our long-cultivated on-site knowledge of food package manufacturing with the innovative technology of AI. We will complete Mikaes to meet the stringent demands of the frontline and contribute to quality improvement and operational efficiency in the food industry as a whole."
Marushin Corporation URL:https://www.maru-sin.co.jp/

Shuji Omura, Representative Director, Tact Pixel Corporation
We expect our image processing and AI technologies to evolve into more practical solutions when combined with the knowledge of professionals in the field."
Tact Pixel Corporation URL:https://taktpixel.co.jp/

Press Release Reprint