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

2026-08-07

AI Summary

This work addresses the instability of pseudo-labels in unsupervised spacecraft image segmentation caused by variations in text prompts and geometric inaccuracies. To this end, the authors propose a contour-prior two-stage distillation framework that integrates multi-prompt voting to generate stable pseudo-masks. In the first stage, the model learns foreground contours; in the second stage, it refines predictions using signed distance fields, skeleton constraints, and area regularization, augmented by a sample-level gating mechanism to filter out unreliable geometric information. Using SAM-generated pseudo-labels and TinyUNet as the lightweight student network, the method achieves gains of 0.0456 in Image IoU and 0.1380 in Boundary F1 on SpaceSense-Bench, with only 0.263M parameters and approximately 1.1 ms per-image inference time, significantly outperforming existing baselines in both boundary quality and foreground accuracy.

Abstract

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.

Problem

• spacecraft segmentation
• annotation-free
• geometric errors
• pseudo-labels
• foundation models

Innovation

• geometry distillation
• silhouette-first learning
• pseudo-mask refinement
• annotation-free segmentation
• signed-distance-field