[2F1-GS-10] AI application:

[2F1-GS-10-02] Human-in-the-Loop machine learning system for disabled-led AI development

〇Kasumi Ushio¹, Ryusuke Ueki¹, Yutaka Hoshina² (1. Qualtec Co.,Ltd., 2. Sumitomo Electric Industries, Ltd.)

Keywords:Human-in-the-Loop, Active Learning, Annotation, Semantic Segmentation, People with Disabilities

In recent years, there has been a growing trend to actively employ people with intellectual disabilities in annotation work in AI development. However, tasks such as training and evaluating AI models are specialized and complex, making it difficult for people with intellectual disabilities to participate in these tasks.
This study developed a Human-in-the-Loop machine learning system that enables people with intellectual disabilities to independently lead a series of tasks, from annotation to training and evaluation, for developing AI for image segmentation. This system automates specialized and complex procedures such as hyperparameter tuning, and its touch-screen user interface allows users to intuitively correct the AI's inference results. By repeating these processes, people with intellectual disabilities can independently complete AI development.
In the verification experiment in which people with intellectual disabilities participated, all of them completed the AI development tasks and gave a favorable evaluation of the system's operability. This research will expand the employment scope of people with intellectual disabilities and pave the way for them to play an even more active role.