Developed an automated measurement app utilizing AI image recognition to contribute to business cost reduction
AI×R Tailor: Automated Measurement PWA
Challenges
In the traditional rental uniform business, size applications were self-reported by staff at client companies. However, size discrepancies due to human error were common, resulting in significant costs for returns and exchanges, including labor (with a maximum return rate of over 40%).
Furthermore, there were issues with environmental considerations, such as the generation of discarded items due to the need to hold more inventory than necessary in preparation for size exchanges.
Unimate aimed to develop and implement an automated measurement app, believing that creating a system to accurately determine sizes before uniform production would improve services for clients.
Solutions
Initially, native app development was desired, but we proposed development as a multi-device compatible PWA, which does not require downloads from stores. This also led to reductions in the time and effort required for store applications and version support, contributing to lower development and operational costs.
AI domain specialists were assigned to develop the AI engine for accurate measurements. Thorough research was conducted on traditional operations, including interviews with veteran staff who had engaged in manual measurements for many years.
Through repeated discussions with Unimate, we created a matching logic that leverages measurement know-how to derive appropriate clothing sizes for the target individuals. We performed everything from requirements definition to detailed design, exploring the development of an AI engine that met the conditions.
Also, designers from Monster Lab's group company A.C.O. (now integrated with Monster Lab) were assigned to handle the development of tone and manner, UI design, and the production of explanatory illustrations. We achieved a design that pursued usability while meeting the requirements arising from the functions of the AI engine.
Results
Through technical research, we derived a method to "create a 3D model from images and predict actual sizes from it," successfully developing an original AI engine. Repeated validations using measurement data provided by Unimate improved the accuracy of AI image recognition.
The application was built with a simple mechanism to provide feedback on suitable sizes from back and side photos of the size measurement target, along with basic data (height, age, weight, and gender). UI design was also added, considering the photographing and management of large numbers of people, resulting in a product with excellent ease of use.
Monster Lab will continue to conduct ongoing verification experiments. We will continue to provide support aimed at enhancing service value from the perspective of size judgment and fit logic improvement, as well as AI accuracy improvement.
(*1) What is PWA?
PWA is short for Progressive Web Apps. It is a new style of website that can provide a user experience similar to a native app on the web. It offers many benefits, such as no installation required as it starts on a browser, the ability to add icons to the home screen and receive push notifications, faster loading and display speeds, and offline viewing.
