Custom-trained machine learning algorithms for space applications
FeatherApps are custom trained machine learning algorithms for spaced-based applications. The algorithms are designed to function on-orbit, and run on AI computing hardware.
The algorithms can be trained to support a range of applications, including: tracking a satellite docking feature using machine vision; detect and locate nearly space objects; perform object detection on Earth observation images in real-time; etc.
FeatherApps are built based on the customer's requirements and the algorithms are trained using customer defined datasets created from real-data, synthetic data, or a combination of both. To ensure flexibility during mission operations, FeatherApps are capable of being configured and updated on-orbit to meet the changing mission needs of any customer.
Autonomous Rendezvous Proximity Operations and Docking
Use machine vision to track satellites on-orbit, enabling autonomous operations in space
Satellite Inspection and Servicing
Map, locate, and detect nearby objects in real-time to support autonomous control of satellite systems
Earth Observation Image Processing
Perform image classification and object detection onboard
Models can run inferences on satellite images in ~3 milliseconds with 96% accuracy.
The on-board machine vision models are capable of being configured and updated in orbit to meet the changing mission needs of any customer.
Send only relevant information to the ground, reducing bandwidth and providing low-latency insight.