Qlean Dataset Launches 'Japanese Daily Housework Motion Video Dataset'

Visual Bank Co., Ltd. (Minato-ku, Tokyo; CEO Masayuki Nagai) will begin providing the 'Japanese Daily Housework Motion Video Dataset' through its subsidiary Amana Images' AI training data solution, 'Qlean Dataset.' This dataset is optimized for improving visual recognition algorithms in home robotics and training advanced Action Recognition models through video analysis.

The dataset consists of video data and metadata recording daily Japanese chores in realistic residential environments within Japan. It covers not just isolated motions but entire tasks, from delicate manual operations like chopping and plating ingredients to full-body movements like vacuuming, hanging laundry, and folding clothes. By including a diverse range of subjects from youth to the elderly and various clothing variations (with/without aprons, different sleeve lengths), it enables the construction of robust models unaffected by visual noise and changes in the real world.

Filming locations are restricted to actual living spaces such as kitchens, living rooms, and laundry areas. This allows AI to deeply learn 'site-specific visual information' like natural light infiltration and indoor lighting reflections, which are difficult to reproduce in simulation environments. In addition to middle shots providing an overview of the entire action, the dataset includes multi-angle close-ups focused on manual tasks. It serves as practical training data for analyzing Physical Object Manipulation (V-O Interaction) and research into context-aware action prediction.

This data is offered as part of the 'AI Data Recipe' lineup—Qlean Dataset's original AI development data series. It strongly supports AI development geared toward social implementation, from the deployment of next-generation domestic service robots to the digitalization of lifestyle habits through video analysis. Visual Bank and Amana Images will continue to support research and development of AI that accurately understands and analyzes the Japanese living environment by providing dynamic structural data captured in Japanese living spaces.

Overview of the 'Japanese Daily Housework Motion Video Dataset' - **Data Type:** Video - **Subject Attributes:** - Actions: Cooking (cutting/serving), Cleaning, Laundry (hanging/folding), Water-related (washing dishes/hands), Others (flower arrangement, tidying up, etc.) - Attributes: All ages and genders, various clothing (aprons, sleeve variations). - **Format:** mp4 / mov - **Environment:** General Japanese households (kitchen, living room, laundry space, etc.) under natural and indoor lighting. Multi-angle composition from middle shots to close-ups. - **Metadata:** Included

Potential Use Cases - **Research:** Verification of action prediction models in environments mixing first-person and third-person perspectives. - **Industrial:** Development of task execution algorithms for autonomous home robots (path planning, manipulator learning). - **Other Needs:** Optimization of lifestyle monitoring systems for the elderly or those needing care (differentiating daily life from emergencies like falls).

About Qlean Dataset 'Qlean Dataset' is a commercially available AI training data solution provided by Amana Images. It supports various data formats (images, videos, audio, 3D, text) and provides a safe environment for both research and commercial use. It continuously expands its 'AI Data Recipe' lineup through collaborations with data holders and media organizations, assisting in the construction of rights-cleared, legally risk-free AI development environments.