Founder
Yoshihiro Banchi
Completed Doctoral Program in Graduate School of Fundamental Science and Engineering, Waseda University, in 2020.
Ph.D. in Engineering
Founded
Inc. at Center for Entrepreneurship, Waseda University in 2020.
Takashi Kawai
Professor, Faculty of Science and Engineering, Waseda University
Ph.D. in Human Sciences
Certified Professional Ergonomist
Officer, Japan Ergonomics Society
Executive Committee, International Ergonomics Association
Concept
Measuring humans. Creating humans.
From small amounts of physiological and psychological data, we predict and generate human behavior, sensibility, and decision-making.
We provide systems and services - Deep Human Analytics - that reconstruct human characteristics across a wide range of domains, forming a new foundation for value creation.
Estimating populations. Expanding with GANs.
GANs (Generative Adversarial Networks) learn underlying features and generate data that is virtually indistinguishable from the real thing. By optimizing GANs, we work to reproduce the complex and ambiguous qualities that make data truly human-like.
Our Approach
Our approach learns human characteristics and smoothly completes the latent space, enabling the generation of more human-like data.
As shown in the figure below, by expanding data that appears noisy in its original form, previously unseen patterns begin to emerge even from small datasets.
Our Approach
Our approach holds significant potential for a wide range of applications, particularly in ergonomics, where obtaining large amounts of human data has become increasingly difficult in today’s environment.
The figures below illustrate examples of applications that utilize our data augmentation approach.
Applications: Database Integration
Integrating different databases that normally cannot be combined
reveals insights that would otherwise remain unknown.
Applications: Database Completion
Completing databases with many missing values or outliers,
filling in gaps such as incomplete or corrupted entries.
Applications: Database Augmentation
Constructing a digital twin by scaling sample data to the target domain
and extracting key segment features.
