Raster Vision

Raster Vision is an open-source Python library and framework with 2,200+ stars on GitHub for building computer vision models on satellite, aerial and other large geospatial imagery — classification, object detection and semantic segmentation included.

Overview

Deep learning on satellite imagery is awkward: images are enormous, geo-referenced, and split across files and coordinate systems, so generic vision tooling breaks down quickly. Raster Vision, developed by Azavea (now Element 84) and still actively maintained in 2026, handles that whole workflow. As a library it provides utilities for reading geo-referenced data, training PyTorch models, running prediction and writing results back out in geo-referenced formats. As a low-code framework it lets users configure complete, repeatable experiments — analyzing training data, chipping imagery, training, evaluating and deploying — through configuration rather than glue code. Built-in support covers the three core geospatial tasks: chip classification, object detection and semantic segmentation, with PyTorch backends throughout. It is the kind of infrastructure remote-sensing teams adopt when a project grows past notebooks into pipelines that must run the same way every time.

Key Features

• Three core tasks – chip classification, object detection and semantic segmentation built in.
• Geospatial native – reads and writes geo-referenced imagery and predictions correctly.
• Full workflow – data analysis, chipping, training, evaluation and prediction in one framework.
• Low-code experiments – configure repeatable pipelines without writing pipeline code.
• PyTorch backends – modern deep learning models under the hood.
• Actively maintained – Apache 2.0 licensed, on PyPI, with stable documentation.

💳 Pricing

Raster Vision is free and open source under the Apache License 2.0. Install it from PyPI and run it on your own machines or cloud instances — there is no paid edition or hosted version.

Who Is It For

Geospatial and remote-sensing teams, earth observation researchers and ML engineers who build vision models over satellite, aerial or drone imagery and need pipelines that respect geography.