Industrial Hot Work Operation Scene Dataset

Description: This dataset contains high-quality synthetic images of industrial hot work operations, including welding, grinding, and cutting tasks that produce visible sparks, with some scenes simulating incomplete personal protective equipment (PPE) usage by workers. It covers diverse industrial plant environments, capturing both safety-compliant and non-compliant scenarios. The images include critical hazard sources such as fire pans, fire extinguishers, acetylene cylinders (placed horizontally or vertically), and oxygen cylinders (positioned at least 5 meters away from acetylene), enabling precise hazard mapping in visual perception models.

The dataset is designed for occupational safety compliance detection, PPE usage recognition, hazardous activity classification, and industrial scene understanding in safety monitoring systems. It supports applications such as real-time worksite risk assessment, worker equipment verification, and hazard zone detection for automated safety enforcement.

All images are captured at 1920×1080 resolution from two viewpoints—1.6 m eye-level and 4 m overhead—at distances of approximately 5–10 meters from the worksite. Lighting conditions are representative of real industrial environments, with sparks and equipment placement clearly visible. Scenes simulate various combinations of workers and equipment, including cases where protective masks, goggles, blue overalls, or safety helmets are missing.

Annotations are provided in object detection format (JSON), including both entity detection and action classification. Label categories cover PPE presence (helmet, goggles, mask), task type (welding, cutting), and hazardous objects (sparks, acetylene cylinder, oxygen cylinder, fire pan, extinguisher). Balanced representation across different PPE states, work types, and hazard object arrangements ensures robust training for safety monitoring AI systems.

By combining photorealistic rendering, human annotation, and realistic industrial hazard simulation, this dataset enables the development of highly accurate safety violation detection models, improves industrial risk monitoring, and supports compliance automation in manufacturing and construction environments.

Keywords: hot work welding grinding cutting PPE detection safety violation industrial hazard monitoring helmet recognition goggles detection

Sector: Industrial Safety / Occupational Health / Computer Vision / Risk Assessment Systems

Metrics: PPE compliance detection accuracy, hazard object detection precision, activity classification recall

Entities: Workers, PPE i tems, welding/cutting tools, sparks, acetylene cylinders, oxygen cylinders, fire extinguishers, fire pans

**Price Range (USD): **4000–20000

Source type: Synthetic data generation from simulation environment, human annotation