About the work
This registered work comprises the source artificial intelligence models developed within the GMVigIA framework, consisting of a set of trained and quantized neural network models specifically designed for on-board Earth Observation (EO) use cases under stringent spaceborne constraints.
The models are optimized for embedded, low-power, and deterministic execution, and have been trained using representative EO datasets to address key operational scenarios, including:
Maritime vessel detection, enabling autonomous identification and localization of ships for surveillance and monitoring applications.
Wildfire detection, supporting early identification of fire outbreaks through onboard analysis of optical and multispectral imagery.
Structure and target detection in SAR imagery, enabling the recognition of man-made structures and relevant features in synthetic aperture radar data.
All models have been subjected to INT8 quantization, preserving inference accuracy while significantly reducing computational load, memory footprint, and power consumption. The quantization-aware design ensures compatibility with edge and on-board AI accelerators, including heterogeneous architectures integrating AI engines, programmable logic, and embedded processors.
The registered material includes the trained model architectures, learned parameters (weights and biases), quantization configurations, and associated preprocessing and postprocessing definitions, constituting an original and protectable body of intellectual property. These models represent a strategic and reusable AI asset, enabling scalable deployment across future EO missions and supporting commercial exploitation through licensing and royalties.
AI Availability Declaration
This work cannot be made available to AI systems.
Creativity declaration
No AI has been used in the creative process of this work
Print work information
Work information
Title GMVigIA Quantized On-Board AI Models for Earth Observation Applications
This registered work comprises the source artificial intelligence models developed within the GMVigIA framework, consisting of a set of trained and quantized neural network models specifically designed for on-board Earth Observation (EO) use cases under stringent spaceborne constraints.
The models are optimized for embedded, low-power, and deterministic execution, and have been trained using representative EO datasets to address key operational scenarios, including:
Maritime vessel detection, enabling autonomous identification and localization of ships for surveillance and monitoring applications.
Wildfire detection, supporting early identification of fire outbreaks through onboard analysis of optical and multispectral imagery.
Structure and target detection in SAR imagery, enabling the recognition of man-made structures and relevant features in synthetic aperture radar data.
All models have been subjected to INT8 quantization, preserving inference accuracy while significantly reducing computational load, memory footprint, and power consumption. The quantization-aware design ensures compatibility with edge and on-board AI accelerators, including heterogeneous architectures integrating AI engines, programmable logic, and embedded processors.
The registered material includes the trained model architectures, learned parameters (weights and biases), quantization configurations, and associated preprocessing and postprocessing definitions, constituting an original and protectable body of intellectual property. These models represent a strategic and reusable AI asset, enabling scalable deployment across future EO missions and supporting commercial exploitation through licensing and royalties.
Work type Source Code
Tags onboard-ai, sar-analysis, quantized-models, earth-observation, edge-ai, int8, machine-learning, gmv, space-applications, wildfire-detection, vessel-detection, artificial-intelligence
-------------------------
Registry info in Safe Creative
Identifier 2512234099487
Entry date Dec 23, 2025, 10:56 AM UTC
License All rights reserved
-------------------------
Copyright registered declarations
Author 100.00 %. Holder GMV Aerospace and Defence SAU. Date Dec 23, 2025.
Information available at https://www.safecreative.org/work/2512234099487-gmvigia-quantized-on-board-ai-models-for-earth-observation-applications
/
Registration: 2512234099487
