HOTOSM African Highway Segmentation Challenge rewards machine-learning developers for road-detection models that help volunteer mappers improve maps for underserved communities.
Funder: Zindi
Due Dates: November 17, 2026 (Start) | December 13, 2026 (Submission close) | December 13, 2026 (Submission selection close) | December 14, 2026 (Leaderboard reveal)
Funding Amounts: $2,500 USD prize pool: $1,200 first place, $800 second place, $500 third place
Summary: A machine-learning competition to detect roads in satellite and aerial imagery and improve mapping for underserved communities.
Key Information: Top-three winners must submit code for verification and assign worldwide copyright in their winning solutions to Zindi.
Hosted on Zindi, this Humanitarian OpenStreetMap Team (HOTOSM) challenge rewards machine-learning models that detect roads in high-resolution satellite and aerial imagery. It addresses missing or outdated road information, particularly in low- and middle-income countries. Models must generalise to regions and imagery sources absent from training data, handling differences in road surfaces, vegetation and image quality while preserving road-network connections.
The challenge supports HOTOSM’s fAIr, an open-source, AI-assisted mapping platform. Predictions will give volunteer mappers draft road networks to review and correct rather than trace from scratch, and winning methods will inform fAIr’s models. Better maps support disaster response, health-service access and infrastructure planning. Openly licensed training data and winning models will also allow other mapping organisations to reuse the methods.