The DEBS 2013 Soccer Monitoring dataset contains high-frequency positional tracking data collected during a complete soccer match using wireless sensors attached to players and embedded in the ball. It records 3D position, velocity, and acceleration at up to 2,000 Hz, generating roughly 15,000 position events per second. The dataset supports research in real-time event processing and sports analytics, including player movement analysis, ball possession, heat maps, and shot detection.

Modality: Spatiotemporal sensor time series / positional tracking data
Recording: One match, two 30-minute halves, 7 players per team
Sampling rate: 200 Hz for player sensors; 2,000 Hz for the ball
Raw dataset size: 2.6 GB (6.0 GB including videos)
Primary use: Real-time sports analytics, event-stream processing

The Warehouse Dataset is a large-scale visual indoor-positioning dataset recorded across 1,320 m² of an industrial warehouse. It contains 464,804 RGB images (640×480) captured by eight cameras, with each image precisely labeled with 3D position and orientation using an optical reference system. It provides dedicated training and testing trajectories designed to evaluate ML-based localization under conditions such as motion artifacts, scale changes, open spaces, shelving areas, and different localization ranges.

Modality: RGB images + spatial pose labels (3D position and orientation/quaternion)
Number of images: 464,804
Complete size: approximately 31.8 GB across the listed training and testing archives
Primary use: Visual indoor localization / ML-based positioning in industrial environments

The Tool Tracking Dataset contains multimodal sensor recordings from four industrial hand tools: an electric screwdriver, pneumatic screwdriver, pneumatic riveting gun, and torque wrench. Accelerometer, gyroscope, magnetometer, and audio data capture typical tool operations as well as anomalous actions such as shaking, hard placement, or component changes. Recordings span approximately 30–60 minutes per tool and contain around 100–200 examples per action type, making the dataset suitable for machine-learning-based activity recognition, anomaly detection, and automated quality assurance in manual industrial processes.

Modality: Multimodal sensor time series — inertial, magnetic-field, and audio data
Tools: 4 industrial hand tools
Complete size: 984 MB
Primary use: Tool/action recognition, anomaly detection, and industrial quality assurance