Answers are grounded in this paper and cite their sources.
Small object detection in remote sensing imagery remains a challenging task due to severe scale variations, weak texture characteristics, and complex background interference. Existing methods often fail to effectively balance receptive field expansion and texture preservation, resulting in degraded feature representations for small objects. To address this issue, we propose a collaborative optimization framework, termed the receptive field–texture preservation network (RTP-Net), which enhances small object detection through three complementary modules: the global–local feature extraction module (GLEM), the adaptive weighted enhancement module (AWEM), and the multiscale attention fusion module (MSAF). In particular, GLEM jointly models global contextual information and fine-grained local textures to alleviate feature degradation caused by scale variation. AWEM further enhances texture preservation and enhances discriminative capability via adaptive feature reweighting. In addition, MSAF improves multiscale feature fusion by incorporating attention mechanisms, which enhance feature consistency while suppressing background interference and reducing misclassification in complex scenes. Extensive experiments conducted on multiple public datasets demonstrate that the proposed method consistently outperforms several state-of-the-art detectors, particularly in small object detection tasks. Notably, RTP-Net achieves 84.1% mAP on the DIOR dataset, 95.3% mAP on the NWPU VHR-10 dataset, and 40.3% mAP on the AI-TOD dataset. These results validate the robustness and generalization capability of the proposed framework across diverse datasets and varying object scales.
No takes yet. Share an insight, caveat, or question.
Zhou et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: