torchgeo and geoai
TorchGeo provides low-level PyTorch primitives (datasets, samplers, transforms) for geospatial machine learning, while GeoAI appears to offer higher-level geospatial AI applications and workflows, making them complementary tools that could be used together in a stack.
About torchgeo
torchgeo/torchgeo
TorchGeo: datasets, samplers, transforms, and pre-trained models for geospatial data
Handles multi-source geospatial imagery with automatic coordinate reference system (CRS) reprojection and resolution alignment across heterogeneous satellite data like Landsat and Sentinel. Provides geospatial-aware samplers that extract patches using geographic coordinates rather than pixel indices, enabling seamless composition of datasets through intersection and union operators. Integrates with PyTorch Lightning for end-to-end training pipelines on remote sensing tasks like semantic segmentation and change detection.
About geoai
opengeos/geoai
GeoAI: Artificial Intelligence for Geospatial Data
Provides end-to-end workflows for satellite imagery analysis by integrating PyTorch, Hugging Face Transformers, and specialized geospatial libraries (TorchChange, segmentation_models.pytorch) with automated chip generation, model training, and inference pipelines. Supports multiple remote sensing data sources and formats (GeoTIFF, JPEG2000, GeoJSON, Shapefile) with seamless QGIS plugin integration and interactive visualization via Leafmap/MapLibre for no-code geospatial AI workflows.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work