Model detects if objects truly exist and where in satellite images
Knowing Whether, Not Just Where: Anchor-Probing Distillation for Open-Set Remote Sensing Visual Grounding
Computer Vision and Pattern Recognition
Summary
Sometimes it's hard to tell if a satellite image actually shows the object you're looking for, not just where it might be if present. The authors developed a system that learns to recognize when an object is really there and when it isn't, by paying attention to stable clues in the image. Their approach uses a teacher-student setup to teach a model how to both locate objects and decide if the evidence is strong enough to say the object is present. This helps the model avoid guessing when it isn't sure, improving accuracy in identifying objects in satellite pictures.
What this means in practice
- •For remote sensing analysts: Identify and localize only objects confidently present in satellite images without false positives for absent objects.
- •For automated surveillance teams: Improve filtering of false alarms by discerning when monitored targets are not visually supported in aerial images.
Authors
Zhihao Chang, Lei Zhai, Shuyuan Yang, Zhixi Feng
Abstract
Open-set visual grounding in remote sensing requires models to determine not only where a queried object is located, but also whether the image provides sufficient evidence for the query at all. Existing closed-set grounding models, however, tend to return plausible boxes even for absent objects, blurring localization confidence with genuine Visual Support. We introduce AnchorProbe-Distill (APD), a teacher-student framework that recasts open-set grounding as Visual Support learning. APD builds on the insight that supported queries should induce coherent spatial anchors, whereas unsupported or mismatched queries tend to produce unstable or misleading localization patterns. APD-T therefore uses the Original Query as a Reference Anchor to interpret query perturbations: Semantic-Preserving Queries retain the annotated target, Near-Category Ambiguous Queries are resolved according to whether they form a reliable alternative anchor or collapse onto the Reference Anchor, and Far-Category Counterfactual Queries provide rejection supervision. This reference-aware reasoning is distilled into APD-S, enabling unified localization and rejection from a single image--query pair. Experiments show that APD enables the resulting model to reject absent targets while also improving its localization of present targets.