From part to whole: Structural semantic knowledge guided airplane detection in remote sensing images

Airplane detection in remote sensing images, which aims to localize and recognize airplanes in aerial or satellite imagery, is of great importance due to its broad applications in both civil and military scenarios. Existing airplane detection methods predominantly rely on holistic object-level features, whereas the use of reusable part-level structural priors remains insufficiently explored. In this paper, we propose a two-stage structural semantic knowledge guided framework for airplane detection, consisting of Semantic Knowledge Generation followed by Part Guided Detection. In the first stage, a Structural Semantic Extraction Module (SSEM) extracts representative part prototypes from airplane samples and forms an offline structural semantic knowledge pool. In the second stage, a Part Semantic Aggregation Module (PSAM) aggregates these prototypes with detection features through an attention-based mechanism to provide part-level structural guidance for feature enhancement. Extensive experiments on DOTA-v1.0-Plane, DOTA-v1.5-Plane, and FAIR1M-v2.0-Plane demonstrate the effectiveness of the proposed framework, especially when airplane appearances are incomplete (e.g., occlusion or boundary truncation). Our code is available at https://github.com/sunhanyang/PartSemantic_PlaneDet.