mlsanityData to 14 Oct 2026

GeoDistill-Refine: Silhouette-First Geometry Distillation for Annotation-Free Spacecraft Segmentation

arXiv 2608.07405 ↗First posted 7 Aug 2026cs.CV cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Foundation segmentation models can provide supervision for spacecraft imagery without manual training masks, but their predictions vary with textual prompts and may contain geometric errors that are amplified during distillation. This paper presents GeoDistill-Refine, a two-stage framework that transfers offline SAM 3 pseudo-masks to a compact segmentation network. Six fixed prompts are fused by an unweighted 50% vote to stabilize the teacher output. The student first learns the foreground silhouette and is then refined with signed-distance-field, skeleton, and area objectives derived from the pseudo-mask. A sample-level gate, computed from prompt agreement, the valid-prompt ratio, and pseudo-mask area plausibility, reduces the influence of unreliable pseudo-geometry. On the SpaceSense-Bench HJM lockbox set, GeoDistill-Refine improves Image IoU and Boundary F1 by 0.0456 and 0.1380, respectively, over a plain pseudo-label student. External evaluations on the SPEED+ Lightbox and Sunlamp domains and on TANGO show competitive regional overlap together with gains in boundary quality or foreground precision. The deployed TinyUNet contains 0.263 M parameters and requires approximately 1.1 ms per image on an RTX 4090; SAM 3 pseudo-mask construction and the auxiliary geometry branches are used only during training.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 21 of its 31 references that are in the map. First posted on arXiv 7 Aug 2026.

Builds on

21 of 31 references.

References · 21

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  2. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  3. 2023Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detectioncited by 3,332
  4. 2020FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidencecited by 2,085
  5. 2019Searching for MobileNetV3cited by 1,994
  6. 2025SAM 3: Segment Anything with Conceptscited by 906
  7. 2023Segment Everything Everywhere All at Oncecited by 548
  8. 2021Semi-Supervised Semantic Segmentation with Cross Pseudo Supervisioncited by 463
  9. 2020clDice -- A Novel Topology-Preserving Loss Function for Tubular Structure Segmentationcited by 294
  10. 2022Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labelscited by 218
  11. 2018Boundary loss for highly unbalanced segmentationcited by 190
  12. 2019Gated-SCNN: Gated Shape CNNs for Semantic Segmentationcited by 166
9 more
  1. 2019Satellite Pose Estimation Challenge: Dataset, Competition Design and Resultscited by 62
  2. 2021SPEED+: Next-Generation Dataset for Spacecraft Pose Estimation across Domain Gapcited by 57
  3. 2023A Survey on Deep Learning-Based Monocular Spacecraft Pose Estimation: Current State, Limitations and Prospectscited by 30
  4. 2021A Spacecraft Dataset for Detection, Segmentation and Parts Recognitioncited by 23
  5. 2019Distance transform regression for spatially-aware deep semantic segmentationcited by 9
  6. 2026SpaceSense-Bench: A Large-Scale Multi-Modal Benchmark for Spacecraft Perception and Pose Estimationcited by 2
  7. 2026GABI: Geometry-Aware Boundary Integration for Spacecraft Segmentationcited by 1
  8. 2026Segmentation-based Detection for Efficient Multi-Task Spacecraft Perceptioncited by 1
  9. 2026Post-Launch Capability Expansion of Vision-Language Models via Prompting for On-Orbit Spacecraft Inspectioncited by 1

Built on by

Not cited yet.