Sunbird/urban-noise-uganda-61k

Dataset Summary

The urban-noise-uganda-61k dataset consists of audio samples representing urban noise environments. It is designed for tasks such as noise classification, audio tagging, or machine learning applications in sound analysis. The dataset includes two configurations, large and small, with varying sizes of data.

Publication

This dataset is described in the following peer-reviewed data descriptor:

Nsumba, S., Muhanguzi, T., Ouma, E.N. et al. Noise mapping and ambient sound recordings of the urban environment in Uganda. Scientific Data 13, 345 (2026). 🔗 https://doi.org/10.1038/s41597-026-06658-w

Featured in Nature Africa — "Fungus fighting fruit flies, dinosaur tracks, and city soundscapes" (Research Highlight, 12 March 2026) 🌍 https://www.nature.com/articles/d44148-026-00056-5

---

Data Collection

The dataset was collected as part of a large-scale noise mapping initiative across Kampala (five divisions) and Entebbe (four wards), Uganda, in collaboration with city authorities (Kampala Capital City Authority(KCCA) and Entebbe Municipal Council).

Methodology

• Mobile-based pipeline: Field collectors used a custom mobile application to capture short audio recordings at specific georeferenced locations across residential, commercial, and mixed-use urban areas.
• Noise sources captured: Traffic, sirens, outdoor vendors, religious gatherings, nightlife/entertainment venues, and construction sites, among others.
• Per-sample annotations: Each recording was tagged with noise category, Sound Pressure Level (SPL) in decibels, GPS coordinates (latitude, longitude, altitude, accuracy), and a timestamp.
• Audio format: Single-channel (mono), sampled at 16 kHz.
• Noise taxonomy: Categories developed in consultation with city authorities to reflect noise types most relevant to urban planning and noise complaint resolution in the Ugandan context.

Coverage

City	Areas
Kampala	Five divisions (Central, Kawempe, Makindye, Nakawa, Rubaga)
Entebbe	Four wards (Four wards (Central Kiwafu, Katabi, Kigungu)

Why This Dataset Is Unique

Unlike existing benchmarks such as UrbanSound8K and SONYC-UST, each sample includes precise geospatial metadata, timestamps, and sound pressure level measurements, enabling detailed spatiotemporal noise mapping. It is also the first large-scale urban sound dataset collected in an African city and curated for African urban noise categories.

---

Additional Resources

• Figshare (full dataset + raw files): https://doi.org/10.6084/m9.figshare.30168901
• Real-time noise map (Kampala/Entebbe): https://noise.sunbird.ai
• Sunbird AI Environmental Sensing: https://sunbird.ai/portfolio/environmental-sensing/

---

Citation

If you use this dataset, please cite the data descriptor:

@article{nsumba2026urban,
  title     = {Noise mapping and ambient sound recordings of the urban environment in Uganda},
  author    = {Nsumba, Solomon and Muhanguzi, Tibabwetiza and Ouma, Evelyn Nafula and
               Sekalala, Imran and Bainomugisha, Engineer and Mwebaze, Ernest and Quinn, John},
  journal   = {Scientific Data},
  volume    = {13},
  pages     = {345},
  year      = {2026},
  publisher = {Springer Nature},
  doi       = {10.1038/s41597-026-06658-w}
}

Example Usage

from datasets import load_dataset

# Load the large configuration
large_dataset = load_dataset("Sunbird/urban-noise-uganda-61k", "large")

# Load the small configuration
small_dataset = load_dataset("Sunbird/urban-noise-uganda-61k", "small")