{ "productLinks": [], "slug": "cashb", "featuredImage": "https://cdn.builder.io/api/v1/image/assets%2Ffb3ccc876dd442c6ae31d776377e35db%2F5a117c9062724ad1822429374ad9a610", "publishedDate": 1743660000000, "seo": { "title": "CASHb Data | CASHb: Cashback App  | Monstarlab", "description": "The 'CASHb' app is Japan's first cashback service provided by Cashbee that collects purchase data from receipts. It creates new direct marketing opportunities for consumer goods manufacturers in the food and daily necessities sectors. Cashbee's partner company, Cashbee Data, acquires and utilizes consumer purchase data from receipt images sent by users, providing B2C companies with opportunities to utilize this data. Monstarlab was responsible for improving the image processing technology to efficiently capture and utilize receipt image data as valid data." }, "site": "jp_en", "qoutesBlock": {}, "clientName": "Cashbee", "headerBanner": "https://cdn.builder.io/api/v1/image/assets%2Ffb3ccc876dd442c6ae31d776377e35db%2F5a117c9062724ad1822429374ad9a610", "overview": "The 'CASHb' app is Japan's first cashback service provided by Cashbee that collects purchase data from receipts. It creates new direct marketing opportunities for consumer goods manufacturers in the food and daily necessities sectors. Cashbee's partner company, Cashbee Data, acquires and utilizes consumer purchase data from receipt images sent by users, providing B2C companies with opportunities to utilize this data. Monstarlab was responsible for improving the image processing technology to efficiently capture and utilize receipt image data as valid data.", "qoute": "", "overviewTitle": "CASHb: Cashback App", "industry": "Enterprise", "relatedLinks": {}, "content2": "

Results:

The improvement of image processing technology has enabled high-precision extraction and analysis, significantly reducing the manual verification work that was previously required and leading to increased operational efficiency.

Furthermore, by semantically analyzing the character data collected through natural language processing, we have succeeded in enhancing the added value of the data and collecting data that can be used for marketing and other purposes.

We have now moved on to the phase of collecting even more data and training the system to improve accuracy.

", "content": "

Challenges:

Considering the social situation where digital technologies such as AI and big data are advancing, Cashbee Data predicted that the value of information assets would increase. They believed it was important to bring data utilization in Japanese business to global standards and create a data business with new value.

The company was an early adopter of the business of collecting purchase data from receipt images using Google's OCR. However, while image data was being collected, a mechanism for accurate data collection was necessary to process the text obtained by OCR into data that could be used for business purposes.

Furthermore, the information processing that had previously failed to correctly read data from images was being done manually, and the associated labor time and cost reduction had become an issue.

The development of high-precision image processing technology was urgently needed, and external partners were required to have specialized knowledge of AI image recognition and high app development expertise.

Solutions:

Monstarlab began by analyzing the existing image recognition methods. When comparing the collected data with actual receipts, it became clear that the quality of the collected receipt images varied, and the text could not be read sufficiently. Therefore, we proposed improving the image processing technology and enhancing the AI's image recognition accuracy.

On the development side, we researched and repeatedly improved extraction techniques to ensure recognition of images from any angle. Character recognition on a line-by-line basis was particularly challenging. Because the characters written on the receipt were not recognized as a single line depending on the image angle, we assigned experienced AI engineers to research methods for extracting character information.

Furthermore, we embarked on developing a natural language processing AI engine to semantically analyze the collected character data and utilize it as marketing data for the retail industry. This involved extracting the characters written on the receipt and then identifying the product by comparing the written product name with the closest product name using the JAN code.\"

", "title": "Efficient receipt data extraction and identification by improving image processing technology" }