Search papers, labs, and topics across Lattice.
The paper introduces Rotated Multi-Kernel RetinaNet (RMK RetinaNet) to improve rotated object detection in remote sensing imagery by addressing issues with receptive field utilization, feature fusion, and angle regression. The model incorporates a Multi-Scale Kernel (MSK) Block, a Multi-Directional Contextual Anchor Attention (MDCAA) mechanism, a Bottom-up Path, and an Euler Angle Encoding Module (EAEM). Experiments on DOTA-v1.0, HRSC2016, and UCAS-AOD demonstrate that RMK RetinaNet achieves state-of-the-art performance and robustness in multi-scale and multi-orientation scenarios.
Achieve state-of-the-art rotated object detection in remote sensing by adaptively tuning receptive fields and stabilizing angle regression.
Rotated object detection in remote sensing imagery is hindered by three major bottlenecks: non-adaptive receptive field utilization, inadequate long-range multi-scale feature fusion, and discontinuities in angle regression. To address these issues, we propose Rotated Multi-Kernel RetinaNet (RMK RetinaNet). First, we design a Multi-Scale Kernel (MSK) Block to strengthen adaptive multi-scale feature extraction. Second, we incorporate a Multi-Directional Contextual Anchor Attention (MDCAA) mechanism into the feature pyramid to enhance contextual modeling across scales and orientations. Third, we introduce a Bottom-up Path to preserve fine-grained spatial details that are often degraded during downsampling. Finally, we develop an Euler Angle Encoding Module (EAEM) to enable continuous and stable angle regression. Extensive experiments on DOTA-v1.0, HRSC2016, and UCAS-AOD show that RMK RetinaNet achieves performance comparable to state-of-the-art rotated object detectors while improving robustness in multi-scale and multi-orientation scenarios.
