a Python package that allows users to transparently use heterogenous geospatial data in PyTorch-based deep learning pipelines. Specifically, TorchGeo provides:
data loaders for common geospatial datasets from the literature
data loaders for combining arbitrary geospatial raster and vector data layers with the ability to sample pixel-aligned patches on-the-fly
augmentations that are appropriate for multispectral imagery and geospatial data
data samplers appropriate for geospatial data
pre-trained models for many common remotely sensed imagery sources
https://github.com/microsoft/torchgeo
a Python package that allows users to transparently use heterogenous geospatial data in PyTorch-based deep learning pipelines. Specifically, TorchGeo provides: