Building and Establishment Automated Mapper (BEAM)
BEAM is an innovative AI/ML tool designed to map building footprints, providing accurate and up-to-date information that helps guide urban development planning, particularly in underserved areas and informal settlements. BEAM's machine learning algorithm uses high-resolution arial or satellite imagery to detect, visualize, and geo-reference infrastructures, transforming complex imagery into actionable datasets.
BEAM can support local governments in maintaining up-to-date spatial data, identifying areas for upgrading and investment, and improving building detection and mapping processes. The tool has already been deployed by multiple local and national government entities across Africa and Central America. Its user-friendly interface enables integration into existing urban planning and data workflows and supports application across different cities and contexts.
Through this Open Call, potential partners are invited to deploy and/or adapt BEAM to a new context and explore its application in addressing locally identified urban challenges. In collaboration with UNITAC, selected partners will adapt and further develop the tool using relevant local data and imagery to produce accurate and context-specific results and contribute to more inclusive and sustainable urban development.
Requirements
1. General Partner Requirements
For applications involving collaboration between governmental and non-governmental entities, the following partnership arrangements are accepted:
- Consortium: A consortium of governmental and non-governmental entities, for example involving a civil society organisation, urban planning lab, or similar organisation.
- Cooperation agreement: A formal cooperation between a governmental and a non-governmental entity, documented through an enclosed Letter of Intent.
The local partner is expected to ensure:
- Local stakeholder engagement: Established connections with relevant stakeholders and the ability to facilitate access to the community data required for the project.
Capacity to use the tool: Relevant staff and technical teams are available and prepared to adopt and actively use the tool in participation, planning, policy, or implementation processes.
2. Resource Requirements
The local partner is expected to allocate or secure the following resources and capacities:
- Project coordination: A dedicated focal point for approximately 6 months to support project coordination, data collection, user testing, and research alongside the UNITAC team.
- Data science and IT support: A local data scientist or IT focal point for approximately 6 months to support tool development alongside the UNITAC team.
- Coordination: A designated focal point to coordinate activities, facilitate stakeholder engagement, support communication between project partners, and oversee local implementation.
- Technical capacity: A local team with the capacity and relevant expertise to support the development and deployment of the tool, including GIS, remote sensing, and machine learning (ML) expertise.
- Implementation planning: A technical implementation plan outlining how the tool will be deployed and integrated locally.
Long-term maintenance: Sufficient capacity and commitment to operate and maintain the tool following implementation.
3. Technical Requirements
Based on our experience with previous implementations of BEAM, we recommend the following technical requirements from the local partner for successful implementation.
Inference server (CPU, production API/UI)
- Operating system: Ubuntu 26.04 LTS
- CPU: Minimum 8 vCPUs; 12 vCPUs recommended
- RAM: Minimum 16 GB; 24 GB recommended
- Storage: Minimum 128 GB SSD; 256 GB SSD recommended
Training and experimentation server (GPU, optional for model retraining)
- Operating system: Ubuntu 26.04 LTS
- CPU: Minimum 8 vCPUs; 12 vCPUs recommended
- RAM: Minimum 16 GB; 24 GB recommended
- Storage: Minimum 512 GB SSD; 1 TB SSD recommended
- GPU: Up to 4 × Tesla T4 or equivalent; resources may be deallocated when not in use
Imagery and training data
- High-resolution imagery: Access to suitable high-resolution aerial or satellite imagery for the proposed area of interest, or a clear pathway to secure such access. Where applicable, a data-sharing agreement or similar arrangement must be in place. Suitable drone imagery may also be used; Sentinel imagery alone is not sufficient.
- Training data: Access to labelled training data, including seed data for the proposed area of interest, or demonstrated capacity to produce the required labelled data.
Domain and security
- Domain: One public domain with 1–2 subdomains available for deployment.
- SSL certificates: Required SSL certificates for the domain and subdomains.
Who is it for?
- Local and national governments
- Urban planners and managers
- Non-governmental organisations (e.g. civil society organisations or urban planning labs
