Mapflow AI for Radio Network planning –

generating 3D objects from 2D satellite images


Mapflow enables the cost-effective deep learning workflow to update 3D maps using only 2D imagery + DEM automatically. The Mapflow workflow engine is designed for large-scale imagery processing with an average speed of 1 min per 4 Megapixel (~ 1 sq. km VHR). Lidar technology and aerial photogrammetry are the most accurate but much more expensive when it comes to the large coverage. Mapflow AI models leverage segmentation of shadows, regression and other ML techniques to perform on a single satellite image with the claimed accuracy of ±3m RMSE in a large and dense urban environment. The results can be tuned to even better accuracy with some manual mapping edits.

Apply construction detection to identify the recent changes in the urban environment

Construction Detection solution leverages satellite time series to provide reports on the construction & development progress. Optimise your costs for the radio planning by timely mapping of the new buildings around the urban areas

Data streaming

Recent satellite imagery powered by the data streaming services enable instant access to global data with automatic processing (e.g. Maxar SecureWatch). You can also use your own UAV / Satellite images or the data subscription to get it securely connected to our AI platform

GIS and API integration

Our solution has an integration with QGIS providing an immediate workspace for data analysts using free and popular desktop software. Solution developers can start integration using Mapflow public API. The deliverables and API can be customized to meet your workflow requirements whenever you work with familiar professional software or download reports.

Open geodata – Github

Send us a message