uGAP
A unified geospatial AI platform for Africa that aggregates more than 120 African geospatial datasets. It provides a data catalog, a Python SDK and machine learning analysis functions, supporting the full workflow from data discovery to model training.
Timeline (last 90 days)
The uGAP platform currently includes 120 open datasets, covering major domains such as agriculture and land cover, of which about 24 are national-level datasets, with additional regional-level and continental-level data.
The uGAP platform offers three data access methods: direct download; drawing a bounding box on a map for spatially clipped download; and filtering Sentinel, MODIS, and Landsat imagery by time window and cloud cover through an integrated search-and-archive function.
uGAP attaches about 34 metadata variables to each dataset and treats metadata completeness as the foundation of credibility.
The uGAP platform hosts all datasets in their original repositories and does not claim ownership of the data itself, allowing users to trace back to the original source.
uGAP provides a Python SDK that encapsulates all capabilities of the catalog and adds analytical functions as well as machine learning pretrained weights and architectures.
The uGAP team is advancing embedded integration with TorchGeo, with the goal of bringing pretrained weights and geospatial model capabilities directly into the SDK.
Analysis of the 2019 Cyclone Idai flood event in Mozambique using the uGAP SDK showed that about 300 square kilometers were completely inundated and about 33,000 people lived within the flood-affected area.
Among uGAP's 120 datasets, the vast majority come from international satellites such as Landsat, Sentinel, and MODIS, and there is as yet no African indigenous satellite data.
uGAP's community feature allows users to submit data links (URL form only), report missing or corrupted data, and suggest improvements, but users cannot directly upload data and can only suggest data locations.
uGAP lists a license field separately in its metadata, and the platform strictly screens for datasets that allow redistribution and open licensing; for nested datasets spanning multiple countries, if some countries restrict distribution, the team communicates with the data source provider.
1 more timeline events
All Facts (11)
The uGAP platform currently includes 120 open datasets, covering major domains such as agriculture and land cover, of which about 24 are national-level datasets, with additional regional-level and continental-level data.
50%UnverifiedThe uGAP platform offers three data access methods: direct download; drawing a bounding box on a map for spatially clipped download; and filtering Sentinel, MODIS, and Landsat imagery by time window and cloud cover through an integrated search-and-archive function.
50%UnverifieduGAP attaches about 34 metadata variables to each dataset and treats metadata completeness as the foundation of credibility.
50%UnverifiedThe uGAP platform hosts all datasets in their original repositories and does not claim ownership of the data itself, allowing users to trace back to the original source.
50%UnverifieduGAP provides a Python SDK that encapsulates all capabilities of the catalog and adds analytical functions as well as machine learning pretrained weights and architectures.
50%UnverifiedThe uGAP team is advancing embedded integration with TorchGeo, with the goal of bringing pretrained weights and geospatial model capabilities directly into the SDK.
50%UnverifiedAnalysis of the 2019 Cyclone Idai flood event in Mozambique using the uGAP SDK showed that about 300 square kilometers were completely inundated and about 33,000 people lived within the flood-affected area.
50%UnverifiedAmong uGAP's 120 datasets, the vast majority come from international satellites such as Landsat, Sentinel, and MODIS, and there is as yet no African indigenous satellite data.
50%UnverifieduGAP's community feature allows users to submit data links (URL form only), report missing or corrupted data, and suggest improvements, but users cannot directly upload data and can only suggest data locations.
50%UnverifieduGAP lists a license field separately in its metadata, and the platform strictly screens for datasets that allow redistribution and open licensing; for nested datasets spanning multiple countries, if some countries restrict distribution, the team communicates with the data source provider.
50%UnverifiedThe uGAP roadmap includes: expanding the metadata system to enhance interoperability (building on the CryoCloud metadata developed by the NASA Middle Commons community with geospatial extensions); continuing to advance machine learning capabilities in the SDK; and acquiring more labeled datasets.
50%