Saved in:
| Hovedforfatter: | |
|---|---|
| Format: | Recurso digital |
| Sprog: | engelsk |
| Udgivet: |
Zenodo
2026
|
| Fag: | |
| Online adgang: | https://doi.org/10.5281/zenodo.20053633 |
| Tags: |
Tilføj Tag
Ingen Tags, Vær først til at tagge denne postø!
|
Indholdsfortegnelse:
- <p>-Updated to new version with corrected co-occurrence interpretation, OOF protocol, and 500 m external validation<br><br>This repository contains the complete source code for the end-to-end urban noise <br>monitoring pipeline described in the manuscript "An End-to-End Deep Learning Pipeline <br>for Urban Noise Monitoring with Spatiotemporal Validation" (under review at Scientific Reports).</p> <p>The pipeline extends from raw data collection (SONYC-UST v2.3 audio + NYC 311 complaint records) <br>through CNN14-based multi-label classification, calibration analysis, TorchScript deployment, <br>spatiotemporal mapping, and external corroboration analysis.</p> <ul> <li>Contents <ul> <li>15 Jupyter notebooks, including the manuscript-reproduction workflow and one legacy data-acquisition notebook retained for transparency</li> <li>README.md with detailed usage instructions, hardware requirements, and reproducibility settings</li> <li>requirements.txt with fully pinned dependency versions</li> </ul> </li> <li>Hardware Tested<br>NVIDIA GeForce RTX 3060 (12 GB VRAM), Python 3.11, PyTorch 2.5.1 with CUDA 12.1.</li> <li>Dataset Access <ul> <li>SONYC-UST v2.3: https://zenodo.org/record/3966543</li> <li>NYC 311 Noise Complaints: NYC Open Data Portal (dataset 76ig-c548)</li> </ul> </li> <li>Citation<br>Please refer to the README.md for citation details. The full manuscript citation <br>will be updated upon publication.</li> <li>License<br>MIT License</li> </ul>