GIR: Geometry and Impulse Response Dataset for Computational Acoustics

The GIR dataset from ETH Zurich contains over 900,000 Impulse Responses (IRs) to study the relationship between geometry and acoustic parameters. Acoustic measurements were taken at 2,952 positions on 312 distinct, 3D-printed surfaces using an automated robotic process. The dataset is released under the GNU General Public License v3.0 with enriched metadata in JSON-LD format.

Use Cases

• Training machine learning models to predict acoustic parameters from geometric surface properties.
• Simulating room acoustics and reverberation effects based on measured impulse responses.
• Validating computational models of sound propagation and scattering in complex environments.
• Studying the relationship between object geometry and its resulting acoustic signature.

Strengths

• Large scale with over 900,000 individual impulse response measurements.
• High consistency ensured by using 3D-printed surfaces and an automated robotic measurement process.
• Detailed provenance and metadata tracked using the Renku platform and Schema.org recommendations.

Limitations

• Specific column names and data schema are not detailed in the provided descriptions.
• The exact dataset size in gigabytes and the precise row count per file are unknown.
• The 'last updated' date is not provided, making freshness difficult to assess.

Provenance

Source

ETH Zurich (Achilleas Xydis)

Collection Method

Acoustic measurements acquired at 2,952 positions on 312 3D-printed surfaces using two robots.