← Back to the feed
Space & aviationResearch findings

Pose Estimation from Real Images Without CAD Models

Original published 2026.05.19 19:37 KSTAdded to Blackdeck 2026.11.22 23:09 KST

What has changed?

Accurate spacecraft pose estimators can be trained using only a few real images without CAD models

Key points

  1. Pose estimators can be trained without CAD-rendered images or synthetic datasets
  2. NeRF-based augmentation generates diverse, geometrically consistent viewpoints and lighting conditions
  3. Accurate pose estimation is achieved using only 25 to 400 real images, even under severe illumination changes

Connect it to your work

No need for CAD models or large synthetic datasets; relies solely on real-world images

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivCAD-Free Learning of Spacecraft Pose Estimators via NeRF-Based Augmentations