AIIC AI Intelligence Centre

SOURCE-LINKED INTELLIGENCE

HGSQ: Heatmap-Guided Sparse Query Detector for Real-Time Aerial Small Object Detection

arXiv · AI, language, vision and robotics · article · Sep 10, 2026 · UTC

Real-time aerial small object detection is an important visual signal and image processing problem, requiring a detector to preserve fine-grained localization while avoiding redundant computation on large background regions. This paper focuses on this deployment-oriented aerial/UAV setting rather than claiming a universal detector for all object detection scenarios. Existing Transformer-based detectors provide strong global modeling, but their dense query initialization and multi-layer decoder still spend substantial computation on background tokens, which is inefficient when small objects occ

Read original source ↗ Open in workspace

recordType
paper
region
Global

Evidence & attribution

First collected: 2026-09-20T19:02:05.452Z. This is not the publication date.