#Japan #Tokyo #Qlean Dataset #Visual Bank #AI Learning Data

Visual Bank Introduces Qlean Dataset for AI Development

Visual Bank Inc., located in Shibuya, Tokyo and led by CEO Masayuki Nagai, is expanding its dataset offerings through its subsidiary, Amana Images. They have announced the launch of the 'Qlean Dataset', an AI learning data solution catering to various research and commercial AI development demands. The latest addition to their lineup is the 'Abnormal Worker Behavior Image Dataset,' aimed at enhancing their AI data recipe offerings.

Overview of the Qlean Dataset

The `Qlean Dataset` comprises original data sets available for commercial use, allowing flexible combinations based on application requirements, accuracy, and delivery times. These data sets include some annotated and unannotated data, with options for customized creation and expansion to meet specific needs. Through partnerships with companies like Chiba Lotte Marines and Toyo Keizai Inc., alongside global networks and new data acquisitions, the dataset variety is continually growing. This development significantly reduces the burden of data collection and preparation in AI development environments, contributing to faster project timelines.

Details of the Abnormal Worker Behavior Dataset

• - Data Type: Images
• - Subject Attributes: 10 males in their 20s
• - Data Volume: 100GB
• - Number of Data Points: 15,857
• - Data Format: JPEG

Sample Details:

• - Equipment Used: Digital still cameras
• - Shooting Distance & Angles: 2.5m and 5m at an overhead angle
• - Subject Movements: Crouching, leaning, falling, pointing, smoking, standing, sitting, etc.
• - Sample Image URL: View Sample

Applications of the Dataset

1. Action Recognition and Model Training

This dataset encompasses everyday actions like standing and sitting, extending to abnormal behaviors like falling or leaning. It is highly suitable for training behavior recognition and posture estimation models.

2. Heat Stroke and Health Condition Monitoring

This dataset can help in systems designed to detect behaviors such as crouching or falling, thus facilitating early identification of heat strokes or health issues. Its practical implementation can significantly enhance safety and accident prevention in factories and construction sites.

3. Detection of Hazardous Behaviors

Real-time alerts for actions that compromise work efficiency and safety, such as smoking or leaning, can be developed using this dataset. Integration with surveillance cameras and IoT sensors can further strengthen accident prevention measures.

4. Action Evaluation and Scoring

By categorizing and assessing normal versus abnormal behaviors, companies can apply the findings to monitor work performance and evaluate behavioral risks. This can support talent development and safety education initiatives.

5. Multi-Angle Data for Enhanced Accuracy

Captured from different distances and angles, this dataset allows the development of models that closely resemble real surveillance environments. The multi-angle approach is expected to improve operational accuracy.

6. VR/AR Simulation and Instruction

Images depicting abnormal behaviors can be employed to develop educational materials in VR/AR settings, aiding in safety training programs and simulations of hazardous situations.

Key Features of Qlean Dataset

The Qlean Dataset is structured to support both research and commercial applications, with necessary consents obtained from all subjects complying with privacy policies across various jurisdictions. The unique 'AI Data Recipe' mechanism shortens investment timelines while maximizing the return on investment during data procurement. Additionally, for unique datasets not included in 'AI Data Recipe,' tailored data creation is possible to meet specific requirements.

Academic Support Program

In a further effort to support academia, Visual Bank has launched a program offering datasets free of charge to universities, research institutions, and non-profit tech developments. Notably, the program includes over 500,000 data points across 80 different types including images, audio, video, and text — addressing the critical need for high-quality, legally compliant learning data in research settings. More details can be found here.

Visual Bank's mission is to develop a next-generation data infrastructure that maximizes AI development capabilities and unlocks the potential of all data types. Their 100% subsidiary, Amana Images, offers various AI learning data solutions alongside tools like 'THE PEN,' aimed at supporting creators. They are also recognized as a participant in the national research and development program 'GENIAC,' accelerating their efforts towards societal implementation.

For more information visit:
Visual Bank
Amana Images