An open source Python library and framework for building computer vision models on satellite, aerial, and large imagery sets.
Raster Vision is an open-source Python library and framework for building and deploying computer vision models on satellite, aerial, and other large-scale geospatial imagery. It solves the problem of applying deep learning to geo-referenced data by providing specialized tools for reading, processing, and outputting predictions in geospatial formats, supporting tasks like chip classification, object detection, and semantic segmentation.
Geospatial analysts, data scientists, and developers working with satellite or aerial imagery who need to implement deep learning workflows, as well as organizations seeking reproducible and scalable analysis of large imagery sets.
Developers choose Raster Vision because it offers a complete, specialized toolkit for geospatial deep learning, combining the flexibility of a library with the ease of a low-code framework, and includes built-in cloud integration for scalable processing.
An open source library and framework for deep learning on satellite and aerial imagery.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Provides a full suite of utilities for reading, processing, and writing geo-referenced imagery and predictions, as highlighted in the README's library description, which simplifies working with formats like satellite data.
Enables users to configure and execute repeatable machine learning pipelines without deep expertise, allowing quick setup through configuration files rather than extensive coding.
Has built-in support for AWS Batch and SageMaker, facilitating scalable processing of large imagery sets directly from the framework, as mentioned in the cloud integration features.
Manages the entire pipeline from training data analysis to model bundling, reducing manual steps and ensuring reproducibility for geospatial deep learning tasks.
Cloud integration is primarily focused on AWS services, which may not suit teams using other cloud providers like Google Cloud or Azure, limiting flexibility.
Requires managing dependencies, Docker images, and running setup scripts, which can be cumbersome for quick prototyping or teams without strong DevOps experience.
Supports only chip classification, object detection, and semantic segmentation with a PyTorch backend, lacking built-in options for other tasks like instance segmentation or alternative frameworks like TensorFlow.