Answers are grounded in this paper and cite their sources.
Object detection in remote sensing imagery is a challenging task in computer vision. The main challenge lies in the significant semantic gap between low‐level and high‐level features in deep networks. To address these issues, this paper proposes a context‐aware semantic alignment detection transformer (CAS‐DETR). Specifically, we address the loss of structural and textural information in deep features by designing an adaptive multiscale context block (AMCB), which enhances high‐level semantic representations through effective multiscale context modeling. Subsequently, to bridge the semantic gap between shallow and deep features, we propose a local‐contextual global fusion module (LGFM) that integrates fine‐grained spatial textures with global semantic context, thereby improving the robustness of multilevel feature representations. CAS‐DETR is evaluated on the VisDrone2019 and SIMD datasets, two established benchmarks that respectively represent the challenges of dense object distribution in aerial imagery and complex background interference in ground‐level scenarios, demonstrating the broad effectiveness of the proposed approach. CAS‐DETR maintains a efficient architecture with only 17.4 million parameters on the VisDrone2019 and SIMD datasets, reducing the model size by approximately 13.4% compared to the baseline, while delivering superior detection performance.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: