#Japan #Tokyo #AI Training #Qlean Dataset #Visual Bank

Introduction

Visual Bank Inc., headquartered in Shibuya, Tokyo, under the leadership of CEO Masayuki Nagai, has recently announced the launch of its Qlean Dataset focused on fall detection videos involving factory workers. This addition aims to enrich the existing AI training data solutions offered through its subsidiary, Amana Images. The Qlean Dataset is tailored for use in both research and commercial AI development, making it an essential resource in the growing field of AI technology.

What is the Qlean Dataset?

The Qlean Dataset represents a unique collection of original data sets that are available for commercial utilization. It features a flexible structure that allows users to combine data materials based on their applications, precision requirements, and timelines. This dataset not only includes pre-annotated data but also offers the possibility to customize datasets according to specific user requirements.

Partnerships with notable organizations such as Chiba Lotte Marines and Toyo Keizai Inc. help expand the dataset's reach and capabilities, ultimately reducing the burden of data collection and organization in AI development processes. This collaborative approach facilitates faster development cycles and more effective AI solutions.

Overview of the Factory Worker Fall Detection Video Dataset

The newly added dataset contains videos demonstrating various fall scenarios, where four individuals (three men and one woman) were filmed in a factory environment using an IP camera. The dataset comes in the MP4 format with 512 files totaling approximately 974.2 MB in size. The videos encapsulate different types of fall patterns, captured in sanitized work attire of three colors, providing a comprehensive array of scenarios.

• - Data Overview:

- Contents: Videos of factory workers experiencing various fall patterns.
- Total Files: 512
- Data Size: 974.2 MB
- Format: mp4

Use Case Scenarios

The fall detection video dataset enables the training of diverse models for detecting falls from different angles, walking speeds (e.g., brisk walking, slow walking, and stationary), and various fall directions. The ability to classify fall movements enables the development of more sophisticated fall detection algorithms.

Training for Anomalous Behavior Detection

This dataset is invaluable for training AI systems in classifying anomalous behaviors by cataloging pre-fall movements, speeds, and fall directions. Such training data can significantly enhance the performance of AI in identifying abnormal actions based on categorized patterns.

Multiple Camera Angles for Validation

Moreover, videos shot from various camera angles and heights allow for simulation and validation of detection performance under different surveillance setups. This pre-evaluation provides essential insights for application in real-world environments, benefiting proof of concept (PoC) initiatives before full deployment.

Behavior Analysis in Factory Environments

The dataset also includes variations such as height, weight, and clothing color within the model group, enriching the data for analyzing variations in detection results due to human differences and environmental factors. An understanding of these variations aids in refining models to minimize false positives.

Efficient Development with Structured Data

A key feature of the Qlean Dataset is its structured and labeled data, which simplifies the development process by embedding shooting conditions and fall patterns in each video. This facilitates automatic classification, evaluation, and annotation, aiding the initial phases of AI model development.

Features of Qlean Dataset

1. Comprehensive and Compliant: The datasets are collected with consent from all subjects, ensuring compliance with various privacy policies and regulations across countries, allowing for safe utilization in research and commercial projects.
2. Fast and Cost-Effective: By offering data through a unique recipe format, Qlean Dataset enables users to procure data with minimized initial investments while maximizing returns on investment (ROI).
3. Customization Available: If there are specific data requirements that are not captured within the existing dataset, Qlean Dataset can create customized data sets tailored to individual needs, showcasing their capacity for high-level data uniqueness.

Conclusion

Visual Bank continues to pioneer innovative data solutions, and the Qlean Dataset stands as a testament to their commitment to providing enhanced support for AI development frameworks. Interested parties are encouraged to explore further through Qlean Dataset’s official site and reach out for inquiries through their contact form.

Upcoming Events

In addition to this launch, Visual Bank is hosting an online seminar on August 20, 2025, discussing copyright risks surrounding generative AI, led by attorneys from ZeLo and Qlean Dataset's representatives. For those interested in legal aspects related to AI development, this initiative is a great opportunity for deeper understanding.

Date: August 20, 2025, 14:00-15:30
Format: Online (Free)
Registration: Seminar Registration

Visual Bank is on a mission to unlock the possibilities of data and build next-gen data infrastructures that amplify AI development capabilities.