UrbanIoT-Anomaly - Multimodal Smart City Dataset

Dataset Overview

The UrbanIoT-Anomaly dataset simulates a smart city environment by capturing synchronized multimodal data from distributed urban sensors and surveillance cameras. Published in August 2025, it supports advanced research in urban analytics, edge computing, and real-time anomaly detection systems.

Multimodal Data Sources

The dataset integrates environmental sensor network data (temperature in °C for urban heat island monitoring, humidity percentage for air quality analysis, gas levels in ppm for pollution detection, vibration in Hz for infrastructure health, noise levels in dB for acoustic environment, and motion detection for activity tracking) and surveillance visual data (image frames, person count for crowd density, and unusual activity flags).

Context-Aware Annotations

Each data instance includes precise timestamps for sensor-image synchronization, location IDs providing spatial context (street, park, intersection, public square), and binary anomaly labels (0 = normal, 1 = anomalous). The dataset contains sensor_data.csv with time-series measurements, image_metadata.csv with surveillance frame information, and 1002 image files.

Applications

Supports urban anomaly detection (environmental hazards, crowd management, infrastructure failure prediction), edge computing research with compact features for real-time processing, multimodal fusion combining sensor time-series and image features, and spatial-temporal analysis leveraging location and timestamp information.