SatHound is a solution to allow non-technical users configuring and executing multi-object detection processes over satellite images, based on deep learning technologies providing the following features:
- Two modes of operation in a clean and powerful user interface with Web-GIS capabilities:
- The training mode is used for teaching the system how to recognize new objects or to improve previous trainings with new examples. No more training datasets to be shared! This is a user-driven training mode, allowing to define unlimited detection targets.
- The hound mode is used for searching target objects on different map areas using already acquired knowledge.
- Batch training lets the user to continue using the hound mode while a SatHound improves it cognitive functions in background. Training status can be consulted at any time.
- Models are trained incrementally providing fast trainings and the possibility to roll-back to previous knowledge status in case of a wrong training.
- REST API for integrating with other systems. This lets external systems to scan geographic areas using the hound mode as you would do using the user interface.
- On-premise, Hybrid or Cloud deployment capabilities.
- The scalable architecture lets you scale-out the whole application for high availability or heavy load scenarios
- A local catalogue for hosting the satellite data products used in the searches.