SynerNet: A Synergistic Hierarchical Network for Detail-to-Semantic Enhancement

Infrared Small Target Detection (IRSTD) remains a long-standing challenge in intelligent visual perception due to the tiny scale of targets and interference from complex backgrounds. In real world scenarios such as UAV inspection, traffic monitoring, and low-altitude surveillance, distant objects including humans, aircraft, and ground facilities often appear as low-contrast, textureless infrared targets, posing significant obstacles to all-weather perception and autonomous task execution. To address the issues of detail loss in small targets and background noise contamination in deep semantic spaces, this paper proposes SynerNet, a novel infrared small target detection network designed for UAV platforms. SynerNet introduces a hierarchical framework that progressively integrates multi-level cues. SynerNet achieved an IOU value of 86.06% on a large publicly available infrared small target dataset, providing a reliable solution for robust infrared perception in low altitude applications based on UAVs.