🌍 Awesome African AI Datasets

A curated, community-maintained list of free and open datasets for Artificial Intelligence and Machine Learning projects focused on Africa.

Datasets are tagged by domain, have Kaggle-style metadata, and are verified to be truly free for research or commercial use (license permitting).

✅ Always check individual dataset licenses before use.

---

📂 Table of Contents

• Natural Language Processing (NLP)
• Speech / Voice
• Computer Vision & Wildlife
• Geospatial & Agriculture
• Climate & Weather
• Health & Demographics
• Contribution Guide

---

🗣 Natural Language Processing (NLP)

MasakhaNER

• Description: Named Entity Recognition datasets for multiple African languages.
• Languages: Yoruba, Hausa, Igbo, Swahili, Amharic, Wolof, Kinyarwanda, and more.
• Size: ~20K annotated sentences.
• Samples: Named entities tagged in context.
• Tasks: NER model training, evaluation, transfer learning.
• Source: Masakhane Project.
• Link: https://github.com/masakhane-io/masakhaner
• License: Mixed permissive licenses.
• Last Updated: 2021-05
• Best For: Low-resource NLP research, multilingual NER.

MasakhaPOS

• Description: POS-tagged datasets for African languages.
• Languages: Yoruba, Hausa, Igbo, Swahili, Wolof, etc.
• Size: ~10K sentences.
• Samples: Tokenized and tagged sentences.
• Tasks: POS tagging model development.
• Source: Masakhane Project.
• Link: https://github.com/masakhane-io/masakhapos
• License: CC BY-SA 4.0.
• Last Updated: 2020-11
• Best For: Linguistic modeling & POS benchmarking.

African Storybooks Corpus

• Description: Children's storybooks in multiple African languages.
• Languages: Zulu, Xhosa, Swahili, Amharic, Hausa, etc.
• Size: 3,000+ books.
• Samples: Parallel text in multiple languages.
• Tasks: Machine translation, text generation.
• Source: African Storybook Project.
• Link: https://www.africanstorybook.org
• License: CC BY 4.0.
• Last Updated: 2023-04
• Best For: Multilingual MT, literacy applications.

---

🎙 Speech / Voice

Mozilla Common Voice — African Languages

• Languages: Swahili, Chichewa, Amharic, Luganda, Kinyarwanda, and more.
• Link: https://commonvoice.mozilla.org/en/datasets
• License: CC0.
• Best For: ASR, TTS.

ALFFA Public Yoruba, Hausa & Wolof Speech Corpora

• Link: https://github.com/getalp/ALFFA_PUBLIC
• Best For: Low-resource ASR.

OpenSLR African Corpora

• Link: http://openslr.org

---

📸 Computer Vision & Wildlife

Snapshot Serengeti (LILA)

• Size: ~3.2M images.
• Link: http://lila.science/datasets/snapshot-serengeti

African Wildlife Dataset (Kaggle)

• Link: https://www.kaggle.com/datasets/biancaferreira/african-wildlife

---

🛰 Geospatial & Agriculture

AfriCultuReS Crop Type Dataset

• Link: https://africultures.net/data

Africapolis Urban Data

• Link: https://africapolis.org

---

🌦 Climate & Weather

CHIRPS

• Link: https://www.chc.ucsb.edu/data/chirps

TAHMO Weather Stations

• Link: https://tahmo.org

FEWS NET Africa Rainfall Estimates

• Link: https://fews.net

---

🏥 Health & Demographics

DHS Program — African Countries

• Link: https://dhsprogram.com

WHO African Health Observatory Data

• Link: https://aho.afro.who.int

Global Health Observatory — Africa

• Link: https://www.who.int/data/gho

---

🤝 Contribution Guide

We welcome pull requests to add, update, or improve dataset entries.

Steps:

1. Fork this repository.
2. Add your dataset entry in the correct section using the template below.

Dataset Entry Template

### Dataset Name
![Domain](https://img.shields.io/badge/Domain-DOMAINCOLOR) ![License](https://img.shields.io/badge/License-LICENSETYPE-green) ![Year](https://img.shields.io/badge/Year-YYYY-orange)
- **Description**: Short description of the dataset.
- **Languages / Geography**: List languages or regions covered.
- **Size**: Approximate size in MB/GB or number of samples.
- **Samples**: Brief description of sample type.
- **Tasks**: List AI/ML tasks supported.
- **Source**: Organization or project name.
- **Link**: [Dataset link](https://example.com)
- **License**: License type.
- **Last Updated**: YYYY-MM.
- **Best For**: Suggested research/application areas.