Core Concepts
The dataset provides a detailed insight into urban sound propagation for research and model development.
Abstract
The Urban Sound Propagation dataset consists of 25,000 data points from 10 cities, each with 2,500 locations. It includes OpenStreetMap images showing urban layouts and sound distribution maps generated using the NoiseModelling v4.0 framework. The dataset is segmented into four subsets: Baseline, Reflection, Diffraction, and Combined tasks to explore different aspects of sound propagation in urban environments. Each data point includes geospatial information like latitude and longitude, sound levels in decibels, temperature, humidity, and unique identifiers. The dataset aims to enhance generative models for predicting sound behavior in complex urban settings.
Stats
Size of Dataset: ~5 GB
Number of Instances: ~100,000
Training Data Points: 19908 x 4
Evaluation Data Points: 3732 x 4
Test Data Points: 1244 x 4