PIXTA Launches 'Worker Image Dataset' for Machine Learning to Support Safety AI in Logistics and Manufacturing

Pixta Inc. (Shibuya-ku, Tokyo; CEO: Daisuke Komata), operator of the digital resource marketplace 'PIXTA,' has announced the release of its 'Worker Image Dataset' as part of its image and video data provision service for machine learning.

This dataset consists of 1,000 carefully selected images featuring diverse subjects (primarily Japanese) in scenarios with high detection difficulty, such as movements around forklifts, occlusion by cargo, falls, and crouching. By providing high-quality, rights-cleared data, PIXTA aims to significantly reduce data collection efforts at development sites and support faster validation and implementation of AI models.

### Dataset Overview - **Name:** Worker Image Dataset - **Quantity:** 1,000 items - **Price:** 99,000 JPY (including tax) - **Annotation:** Available as an optional paid service upon consultation. - **Locations:** Actual warehouses, distribution centers, manufacturing plants, and studios. - **Compositions:** Bird's-eye view, eye-level, and occlusions by shelves or equipment. - **Cast:** Men and women across a wide range of ages (20s to 60s), primarily Japanese. - **Content:** Human actions in factories/warehouses (work, inspection, operation), movements near heavy machinery (forklifts), and posture variations like falling or crouching.

### Key Features - 100% real-life photography materials. - Commercial use allowed. - Permissions for machine learning use obtained from all photographers.

### Intended Applications This dataset is suitable for corporations and research institutions developing AI models for safety management, action recognition, and work analysis in manufacturing and logistics. Potential use cases include:

1. **Visualization of Work Processes and Flow Analysis:** Building models for picking, cargo handling, and line maintenance using diverse angles. 2. **Building Detection Models for Edge Cases:** Training AI to recognize postures or scenes where detection is difficult, such as when parts of the body are hidden by equipment or in complex working positions. 3. **Foundation for Object Detection and Pose Estimation AI:** Serving as base data for understanding human behavior in industrial settings within the field of computer vision.

### Background of Provision PIXTA has received feedback from AI developers in manufacturing and logistics regarding the difficulty of using overseas datasets, which often fail to replicate the specific environments of Japanese worksites. Factors like unique Japanese workwear, equipment, and specific lighting conditions in local warehouses are essential for improving AI accuracy. As AI moves from the PoC phase to actual operation, there is an increasing demand for high-quality additional data that addresses diverse edge cases like falls and blind spots.