GisForWeb
All tools
Raster Vision logo

Raster Vision

Open Source

Open-source Python framework for deep learning on satellite and aerial imagery.

License

Apache-2.0

About

Raster Vision is a Python framework for applying deep learning to satellite, aerial, and other large geospatial imagery. It provides configurable pipelines for chip classification, object detection, and semantic segmentation, with georeferenced raster and vector predictions. Its integrations cover STAC, object storage, and execution on AWS Batch or SageMaker.

Features

  • Supports chip classification and semantic segmentation
  • Builds object detection training pipelines
  • Reads georeferenced raster and vector data
  • Produces georeferenced raster and vector predictions
  • Loads imagery from STAC APIs and object storage
  • Runs pipelines on AWS Batch and SageMaker
  • Defines reusable YAML training pipelines

Details

Type
Library
Pricing
open-source
License
Apache-2.0
API
sdk
Hosting
Self-host

Platforms

linux
macos
docker

Languages

Python

Data formats

GeoTIFF
GeoJSON
STAC