• vision

  • Vision Program

    These are some of the areas we work on:
    -Estimate product attributes for large-scale image datasets
    -Estimate image suitability for advertising products
    -Smooth customer onboarding with eKYC
    -Generate recommendation to create attractive designs using image and text content
    -Create efficient inspection methods using drone captured footage
    PUBLICATIONS
    OPEN POSITIONS

Provide a frictionless image based experience using advanced computer vision research

The mission of vision group, a part of Rakuten Institute of Technology, is to use advanced computer vision research for automation and new service creation. This group consists of several experienced PhD researchers specializing in computer vision domain and motivated to use their research knowledge for driving innovation at Rakuten.

Focus areas include initiatives such as developing an accurate and scalable eKYC (electronic Know Your Customer) solution using optical character recognition and face authentication to improve customer experience by drastically reducing time taken for customer onboarding, generating creative AI solutions to improve the aesthetic quality and conversion rate of advertisements, extracting visual attributes like colors or brands from product images to enrich the product catalog, and analyzing drone-captured images to automate the antenna audits of Rakuten Mobile.

Frictionless experience with images

Images play an essential role for communication in shopping experiences.

Leveraging state-of-the-art computer vision technology, the Vision program builds novel systems to improve such experiences. In addition, pictures such as drivers' licenses are often used in a critical identification process called Know Your Customer (KYC). As one of the largest Fintech companies in Japan, removing friction from identification processes is critical for the customer experience.

We are developing robust e-KYC algorithms to address these challenges and create a multi-modal approach combining different types of input data.

OTHER RESEARCH AREAS

LEARN MORE