# Fast Learning of Non-Cooperative Spacecraft 3D Models through Primitive Initialization

arXiv: https://arxiv.org/abs/2507.19459 · HTML: https://mlsanity.com/paper/2507.19459/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 14 of its 25 references that are in the map.
- First posted on arXiv 25 Jul 2025.

## Builds on (most cited first, 14 of 14)

- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [3D Gaussian Splatting for Real-Time Radiance Field Rendering](https://mlsanity.com/paper/2308.04079/) (2023)
- [On the Continuity of Rotation Representations in Neural Networks](https://mlsanity.com/paper/1812.07035/) (2018)
- [INeRF: Inverting Neural Radiance Fields for Pose Estimation](https://mlsanity.com/paper/2012.05877/) (2020)
- [Deep Learning for Spacecraft Pose Estimation from Photorealistic Rendering](https://mlsanity.com/paper/1907.04298/) (2019)
- [Robust Multi-Task Learning and Online Refinement for Spacecraft Pose Estimation across Domain Gap](https://mlsanity.com/paper/2203.04275/) (2022)
- [Pose Estimation for Non-Cooperative Spacecraft Rendezvous Using Convolutional Neural Networks](https://mlsanity.com/paper/1809.07238/) (2018)
- [A Spacecraft Dataset for Detection, Segmentation and Parts Recognition](https://mlsanity.com/paper/2106.08186/) (2021)
- [Adaptive Neural Network-based Unscented Kalman Filter for Robust Pose Tracking of Noncooperative Spacecraft](https://mlsanity.com/paper/2206.03796/) (2022)
- [CRISP: Object Pose and Shape Estimation with Test-Time Adaptation](https://mlsanity.com/paper/2412.01052/) (2024)
- [Characterizing Satellite Geometry via Accelerated 3D Gaussian Splatting](https://mlsanity.com/paper/2401.02588/) (2024)
- [Bridging the Domain Gap for Flight-Ready Spaceborne Vision](https://mlsanity.com/paper/2409.11661/) (2024)
- [DreamSat: Towards a General 3D Model for Novel View Synthesis of Space Objects](https://mlsanity.com/paper/2410.05097/) (2024)
- [Training Datasets Generation for Machine Learning: Application to Vision Based Navigation](https://mlsanity.com/paper/2409.11383/) (2024)

## Built on by (most cited first, 5 of 5)

- [Improved 3D Gaussian Splatting of Unknown Spacecraft Structure Using Space Environment Illumination Knowledge](https://mlsanity.com/paper/2512.23998/) (2025)
- [DreamSat-2.0: Towards a General Single-View Asteroid 3D Reconstruction](https://mlsanity.com/paper/2508.01079/) (2025)
- [Dynamic Scene 3D Reconstruction of an Uncooperative Resident Space Object](https://mlsanity.com/paper/2509.07932/) (2025)
- [From Images2Mesh: A 3D Surface Reconstruction Pipeline for Non-Cooperative Space Objects](https://mlsanity.com/paper/2605.00147/) (2026)
- [NeRF-based Spacecraft Reconstruction from Monocular Imagery Under Illumination Variability and Pose Uncertainty](https://mlsanity.com/paper/2605.18447/) (2026)

## Abstract (arXiv)

The advent of novel view synthesis techniques such as NeRF and 3D Gaussian Splatting (3DGS) has enabled learning precise 3D models only from posed monocular images. Although these methods are attractive, they hold two major limitations that prevent their use in space applications: they require poses during training, and have high computational cost at training and inference. To address these limitations, this work contributes: (1) a Convolutional Neural Network (CNN) based primitive initializer for 3DGS using monocular images; (2) a pipeline capable of training with noisy or implicit pose estimates; and (3) and analysis of initialization variants that reduce the training cost of precise 3D models. A CNN takes a single image as input and outputs a coarse 3D model represented as an assembly of primitives, along with the target's pose relative to the camera. This assembly of primitives is then used to initialize 3DGS, significantly reducing the number of training iterations and input images needed -- by at least an order of magnitude. For additional flexibility, the CNN component has multiple variants with different pose estimation techniques. This work performs a comparison between these variants, evaluating their effectiveness for downstream 3DGS training under noisy or implicit pose estimates. The results demonstrate that even with imperfect pose supervision, the pipeline is able to learn high-fidelity 3D representations, opening the door for the use of novel view synthesis in space applications.