20 licensed datasets across 6 industries and 50+ curated open-access datasets for African AI research. From healthcare to finance, NLP to computer vision — ethically sourced, human-validated, and built for local relevance and global impact.
No datasets found matching your criteria.
Finding the right dataset for your AI model shouldn’t take weeks. Discover 50+ curated open-access datasets for building AI that works for Africa — spanning NLP, speech, computer vision, and multimodal tasks across 35+ African languages.
| Dataset | Modality | Tasks | Languages | License | Quality | Links |
|---|---|---|---|---|---|---|
African Breast Ultrasound Dataset Annotated breast ultrasound images for cancer screening in Black African women. |
Image | Image ClassificationObject Detection |
— | CC BY-NC 4.0 | ★★★★★ | |
African Voices Speech Dataset Rich audio-text datasets in four major African languages. |
Audio | ASRLanguage Identification |
HausaYorubaIgboNigerian Pidgin |
CC BY 4.0 | ★★★★★ | |
AfriCaption Multi-lingual image captioning benchmark for 20 African languages. |
Multimodal | Image CaptioningVisual Question Answering |
YorubaHausaIgboAmharic+15 |
CC BY 4.0 | ★★★★★ | 🤗 |
AfriDocMT Document-level multi-parallel machine translation corpus covering English and 5 African languages across health and IT news domains. |
Text | Machine Translation |
AmharicHausaSwahiliYoruba+2 |
CC BY-NC-SA 3.0 | ★★★★★ | 🤗 |
AfriQA Cross-lingual open-retrieval question answering benchmark for 10 African languages. |
Text | Question AnsweringInformation Retrieval |
BembaFonHausaIgbo+6 |
CC BY-SA 4.0 | ★★★★★ | 🤗 |
AfriQA Gold Passages Gold Wikipedia passage retrieval corpus for the AfriQA cross-lingual open-retrieval QA benchmark across 10 African languages. |
Text | Information RetrievalQuestion Answering |
BembaFonHausaIgbo+6 |
CC BY-SA 4.0 | ★★★★★ | 🤗 |
AfriSenti-SemEval 2023 Multilingual Twitter sentiment analysis dataset for 14 African languages from SemEval 2023 Task 12. |
Text | Sentiment AnalysisText Classification |
AmharicAlgerian ArabicHausaIgbo+10 |
CC BY 4.0 | ★★★★★ | 🤗 |
AfriSpeech-200 Pan-African accented English speech dataset covering 200 hours from clinical and general domains. |
Audio | ASRAccent Classification |
English (African accents) |
CC BY 4.0 | ★★★★★ | 🤗 |
ALFFA Speech Corpus Read-speech ASR corpus for five African languages collected in-country by academic partners. |
Audio | ASR |
AmharicSwahiliWolofHausa+1 |
CC BY 4.0 | ★★★★★ | |
Amharic News Text Classification Dataset Amharic news articles scraped from Ethiopian outlets for multi-class topic classification. |
Text | Text ClassificationLanguage Modeling |
Amharic |
MIT | ★★★★★ | 🤗 |
Aya Dataset - Cohere Labs (African subset) Massive multilingual human-annotated instruction-following dataset covering 101 languages including 20+ African. |
Text | Language ModelingQuestion AnsweringSummarisation+2 |
AmharicChichewaHausaIgbo+9 |
Apache 2.0 | ★★★★★ | 🤗 |
Bean Disease Detection Dataset (iBean) Field-captured images of bean leaves for diagnosing common diseases in East Africa. |
Image | Image Classification |
— | CC BY 4.0 | ★★★★★ | 🤗 |
BraTS-Africa MRI brain imaging dataset with expert segmentation for brain tumors in African patients. |
Image | SegmentationImage Classification |
— | CC BY 4.0 | ★★★★★ | |
Cassava Leaf Disease Classification Expert-annotated leaf image dataset for classifying five cassava disease conditions in Ugandan farms. |
Image | Image Classification |
— | CC BY 3.0 IGO | ★★★★★ | 🤗 |
Common Voice Scripted Speech 25.0 — Amharic Crowdsourced Amharic read-speech corpus from Mozilla's Common Voice platform. |
Audio | ASR |
Amharic |
CC0 1.0 | ★★★★★ | |
Common Voice Scripted Speech 25.0 — Hausa Crowdsourced Hausa read-speech corpus from Mozilla's Common Voice initiative. |
Audio | ASR |
Hausa |
CC0 1.0 | ★★★★★ | |
Common Voice Scripted Speech 25.0 — Igbo Crowdsourced Igbo read-speech corpus from Mozilla's Common Voice initiative. |
Audio | ASR |
Igbo |
CC0 1.0 | ★★★★★ | |
Common Voice Scripted Speech 25.0 — Kinyarwanda Crowdsourced Kinyarwanda speech corpus, one of the largest African language datasets on Common Voice. |
Audio | ASR |
Kinyarwanda |
CC0 1.0 | ★★★★★ | |
Common Voice Scripted Speech 25.0 — Luganda Crowdsourced Luganda read-speech corpus from Mozilla's Common Voice platform. |
Audio | ASR |
Luganda |
CC0 1.0 | ★★★★★ | |
Common Voice Scripted Speech 25.0 — Swahili Crowdsourced Swahili read-speech corpus from Mozilla's Common Voice initiative. |
Audio | ASR |
Swahili |
CC0 1.0 | ★★★★★ | |
Common Voice Scripted Speech 25.0 — Tigrinya Crowdsourced Tigrinya read-speech corpus from Mozilla's Common Voice initiative. |
Audio | ASR |
Tigrinya |
CC0 1.0 | ★★★★★ | |
Common Voice Scripted Speech 25.0 — Yoruba Crowdsourced Yoruba read-speech corpus from Mozilla's Common Voice initiative. |
Audio | ASR |
Yoruba |
CC0 1.0 | ★★★★★ | |
FLORES+ African Languages Subset Evaluation benchmark dataset for multilingual machine translation covering 229 languages. |
Text | Machine Translation |
AmharicBembaChokweEwe+23 |
CC BY-SA 4.0 | ★★★★★ | 🤗 |
GhanaNLP Twi and English Parallel Data Twi language NLP resources including parallel text and news data from the GhanaNLP initiative. |
Text | Machine TranslationLanguage ModelingText Classification |
TwiEnglish |
CC BY 4.0 | ★★★★★ | 🤗 |
Google FLEURS — African Languages Subset Few-shot speech benchmark with 20+ African languages for ASR and language identification. |
Audio | ASRLanguage Identification |
AfrikaansAmharicFulaGanda+16 |
CC BY 4.0 | ★★★★★ | 🤗 |
Google Open Buildings ML-derived building footprint dataset covering ~516M structures across Africa and South/Southeast Asia. |
Image | Object DetectionSegmentationGeospatial Analysis |
— | CC BY 4.0 | ★★★★★ | |
GRID3 African Settlement Mapping Dataset High-resolution settlement boundaries and population estimates for 14+ African nations. |
Image | Geospatial AnalysisSegmentation |
— | CC BY 4.0 | ★★★★★ | |
HausaVG Hausa Visual Genome for multi-modal English-to-Hausa machine translation, created by professional translators at BUK. |
Multimodal | Machine TranslationLanguage Modeling |
HausaEnglish |
CC BY-NC-SA 4.0 | ★★★★★ | 🤗 |
HausaVQA Bilingual Hausa–English visual question answering dataset with image–question–answer triplets from HausaNLP. |
Multimodal | Visual Question AnsweringImage Captioning |
HausaEnglish |
CC BY-SA 4.0 | ★★★★★ | 🤗 |
iCassava FGVC 2019 Challenge Dataset Earlier Makerere cassava disease image dataset used in the iCassava 2019 FGVC workshop challenge. |
Image | Image ClassificationFine-Grained Recognition |
— | MIT | ★★★★★ | |
IgboNLP Corpus Gold-standard Igbo NLP resources covering POS tagging, NER, and morphological analysis. |
Text | POS TaggingNERMorphological Analysis |
Igbo |
Apache 2.0 | ★★★★★ | 🤗 |
JW300 African Language Pairs Large parallel corpus from Jehovah's Witnesses publications covering 100+ African language pairs. |
Text | Machine TranslationLanguage Modeling |
SwahiliLingalaShonaZulu+9 |
Unknown | ★★★★★ | 🤗 |
KINNEWS and KIRNEWS Kinyarwanda and Kirundi news classification corpora scraped from regional online news outlets. |
Text | Text ClassificationLanguage Modeling |
KinyarwandaKirundi |
MIT | ★★★★★ | 🤗 |
MAFAND-MT News-domain machine translation benchmark for 21 African languages paired with English and French. |
Text | Machine Translation |
AmharicBambaraGhomalaEwe+19 |
CC BY-NC 4.0 | ★★★★★ | 🤗 |
Makerere Fall Armyworm Crop Dataset Drone and ground-level imagery for detecting fall armyworm infestation in Ugandan maize fields. |
Image | Object DetectionImage Classification |
— | CC BY 4.0 | ★★★★★ | |
Masakhane MT Benchmark 2020 Community-built machine translation benchmark for 38 African language pairs against English. |
Text | Machine Translation |
YorubaHausaSwahiliShona+2 |
Apache 2.0 | ★★★★★ | |
MasakhaNER 2.0 Human-annotated named entity recognition benchmark spanning 20 African languages. |
Text | NER |
BambaraGhomalaEweFon+16 |
Apache 2.0 | ★★★★★ | 🤗 |
MasakhaNEWS News topic classification dataset spanning 16 African languages across 7 categories. |
Text | Text Classification |
AmharicEnglishFrenchHausa+12 |
CC BY-NC 4.0 | ★★★★★ | 🤗 |
MasakhaPOS Part-of-speech tagging dataset covering 20 African languages with Universal Dependencies annotations. |
Text | POS Tagging |
BambaraGhomalaEweFon+16 |
CC BY-NC 4.0 | ★★★★★ | 🤗 |
MASSIVE (African Languages Subset) Multilingual Amazon Slu resource for slot-filling, intent classification, and dialogue tasks covering 52 languages. |
Text | Text ClassificationDialogue |
AfrikaansAmharicSwahili |
CC BY 4.0 | ★★★★★ | 🤗 |
MENYO-20k Multi-domain Yoruba–English parallel corpus of 20,000 human-translated sentence pairs. |
Text | Machine Translation |
YorubaEnglish |
CC BY-NC 4.0 | ★★★★★ | 🤗 |
NaijaSenti Twitter sentiment analysis dataset for four Nigerian languages annotated by native speakers. |
Text | Sentiment Analysis |
YorubaHausaIgboNigerian Pidgin |
CC BY 4.0 | ★★★★★ | 🤗 |
NCHLT Speech Corpus Read-speech corpus covering all 11 official South African languages with phoneme-level transcriptions. |
Audio | ASRTTS |
ZuluXhosaAfrikaansNorthern Sotho (Sepedi)+7 |
CC BY 3.0 | ★★★★★ | 🤗 |
NIH Malaria Parasite Cell Image Dataset Expert-annotated blood smear images for automated malaria detection. |
Image | Image ClassificationObject Detection |
— | Public Domain | ★★★★★ | |
PlantVillage Crop Disease Dataset Expert-labelled plant leaf images covering 26 disease categories across 14 crop types, widely used in African agricultural AI. |
Image | Image Classification |
— | CC BY-SA 3.0 | ★★★★★ | 🤗 |
Radiant MLHub — Rwanda Field Boundary Competition Dataset Sentinel-2 satellite time-series imagery with expert-delineated agricultural field boundaries across Rwanda. |
Image | SegmentationObject DetectionGeospatial Analysis |
— | CC BY 4.0 | ★★★★★ | |
Snapshot Serengeti Millions of camera trap images from Serengeti with crowdsourced wildlife labels. |
Image | Image ClassificationObject Detection |
— | CC BY 4.0 | ★★★★★ | |
Sunbird African Language Technology (SALT) Dataset SALT is a multi-way parallel text and speech corpus of English and six languages widely spoken in Uganda and East Africa. |
Multimodal | Machine TranslationText ClassificationNER |
English (Ugandan accent)LugandaAcholiLugbara+6 |
CC BY 4.0 | ★★★★★ | 🤗 |
WAXAL Google's large-scale multilingual speech corpus spanning 24 African languages for ASR and TTS. |
Audio | ASRTTS |
AcholiAkanAmharicDagbani+20 |
CC BY 4.0 | ★★★★★ | 🤗 |
XL-Sum African Languages Subset BBC-sourced cross-lingual abstractive summarisation dataset covering 10 African languages. |
Text | SummarisationLanguage Modeling |
AmharicHausaIgboKirundi+6 |
CC BY-NC-SA 4.0 | ★★★★★ | 🤗 |
No datasets match your filters. Try adjusting your search.
For most African languages, production-ready speech data doesn't exist at any price. DataLens's on-the-ground contributor network closes that gap. Name the language, dialect, domain, and volume — we launch a collection campaign, trigger guided recording missions with native speakers, run multi-pass QA, and deliver a clean, licensed corpus in weeks.
Campaigns we're actively scoping with partners. Don't see your language, dialect, or domain? We collect those too.
Nigeria's everyday lingua franca — informal, heavily code-switched speech across every region.
Use-Case: Mass-market conversational AI, ASR & voicebots
East Africa's cross-border lingua franca, from Nairobi and Mombasa across the wider region.
Use-Case: Regional ASR, IVR & voice assistants
Tonal South-West Nigerian speech in both formal and street registers.
Use-Case: Nigerian ASR, TTS & contact-centre automation
South-East Nigerian speech with heavy English code-switching.
Use-Case: Nigerian ASR & local-language IVR
Northern Nigeria and the Sahel's dominant language, in broadcast and market registers.
Use-Case: West African ASR & broadcast transcription
Ethiopia's working language, written in Ge'ez script with rich morphology.
Use-Case: Ethiopian ASR, TTS & voice document pipelines
Central Kenya's largest first language, spoken by rural and Nairobi communities alike.
Use-Case: Kenyan multilingual ASR & agri-voice services
Ghana's most widely spoken Akan variety, routinely code-switched with English.
Use-Case: Ghanaian ASR & local-language IVR
Rwanda's universal language, spoken by nearly the entire population.
Use-Case: National ASR, IVR & public-service voicebots
South Africa's largest home language, tonal with click consonants.
Use-Case: SA ASR, TTS & contact-centre automation
Zimbabwe's majority language across Harare and the rural provinces.
Use-Case: Zimbabwean ASR & voice services
The Horn of Africa's largest Cushitic language, written in Latin-script Qubee.
Use-Case: Ethiopian multilingual ASR & IVR
Central Uganda's lingua franca, spanning Kampala urban and rural speech.
Use-Case: Ugandan ASR & local-language IVR
Botswana's national language, also widely spoken across South Africa's North West.
Use-Case: Cross-border ASR & voice services
A Germanic language of Southern Africa with distinct regional accents.
Use-Case: SA ASR, TTS & media transcription
Need a language, dialect, or domain that isn't listed? We scope and staff a custom collection mission with our native-speaker network — same elicitation, QA, and agreement standard.
Generic global datasets fail to capture the nuance, language diversity, and economic realities of African markets. DataLens Africa was built to close that gap — with data that is contextually correct, ethically sourced, and ready for production.
Request access to our full catalog of licensed human-validated African dataset or request a custom data tailored to your project.