aiqbench analysis
This research advances monocular spacecraft pose estimation by leveraging a large synthetic-trained vision foundation model (DINOv3) to achieve state-of-the-art accuracy on non-cooperative targets without using real-world data. The study validates the feasibility of running this 840M parameter architecture on embedded hardware like the Jetson Orin NX, demonstrating that high-fidelity orbital navigation can be achieved through purely synthetic training pipelines.
Vision Foundation Models with Synthetic-Only Training for Monocular Spacecraft Pose Estimation
arXiv:2609.26561v1 Announce Type: cross Abstract: We present an improvement on previous spacecraft pose estimation architectures that results in the lowest published mean rotation errors we know of on the SPEED+ lightbox and sunlamp test sets for a known, non-cooperative spacecraft. By using a previously established heatmap-based pose estimation architecture and adapting a large self-supervised ViT foundation model (DINOv3) in place of the smaller convolutional and ViT encoders of previous work, we show that pose estimation accuracy improves from 300M to 840M parameters with no saturation yet observed. We also evaluate our 840M model on a Jetson Orin NX 16GB, measuring single-pass network inference at 133.8 ms per crop with a board draw of 32.0 W. These measurements demonstrate embedded inference feasibility on a processor family with orbital flight heritage. Our resulting model outperforms previous models across lightbox and sunlamp domains while training only on synthetic data. Our best model, using DINOv3 840M adapted with LoRA as the encoder (rank 64, three-seed ensemble with four-rotation test-time augmentation), results in $1.56^\circ$ mean rotation error on sunlamp and $1.17^\circ$ on lightbox, compared to the previous best mean rotation errors we know of on these test sets, $2.66^\circ$ and $1.75^\circ$ by EagerNet.
Abstract:We present an improvement on previous spacecraft pose estimation architectures that results in the lowest published mean rotation errors we know of on the SPEED+ lightbox and sunlamp test sets for a known, non-cooperative spacecraft. By using a previously established heatmap-based pose estimation architecture and adapting a large self-supervised ViT foundation model (DINOv3) in place of the smaller convolutional and ViT encoders of previous work, we show that pose estimation accuracy improves from 300M to 840M parameters with no saturation yet observed. We also evaluate our 840M model on a Jetson Orin NX 16GB, measuring single-pass network inference at 133.8 ms per crop with a board draw of 32.0 W. These measurements demonstrate embedded inference feasibility on a processor family with orbital flight heritage. Our resulting model outperforms previous models across lightbox and sunlamp domains while training only on synthetic data. Our best model, using DINOv3 840M adapted with LoRA as the encoder (rank 64, three-seed ensemble with four-rotation test-time augmentation), results in $1.56^\circ$ mean rotation error on sunlamp and $1.17^\circ$ on lightbox, compared to the previous best mean rotation errors we know of on these test sets, $2.66^\circ$ and $1.75^\circ$ by EagerNet.
6 pages, 3 figures, 4 tables. A shorter version was accepted to the IROS 2026 Space Robotics Workshop (non-archival)
Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
arXiv:2609.26561 [cs.CV]
(or arXiv:2609.26561v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.26561 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 311 words
Related coverage
Geometry-Consistent Endoscopic Representations for Image-Guided Navigation via Structured Foundation Model Adaptation
arXiv · Jun 17, 2026
NSF-HRPT: Neural Semantic Field meets Hierarchical Risk Perception Tree for Safety-Critical Scenario Assessment
arXiv · Aug 6, 2026
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
NVIDIA Nemotron 3 Nano Omni Powers Multimodal Agent Reasoning in a Single Efficient Open Model
NVIDIA · Apr 28, 2026