GeoCueFormer: Geometry-Guided Wavelet Representation and Prediction-Cued Dual-Stage Decoder for Underwater Semantic Segmentation
本文提出GeoCueFormer，通过几何引导的小波表示和预测提示的双阶段解码器解决水下图像因光线吸收和散射导致的视觉退化问题，提升水下语义分割性能。
0 citationsRead paper
Selected work
Representative Papers
Characterizing Text Branch Sensitivity in Medical Vision-Language Segmentation via Evidence Decoupling
研究通过Evidence Decoupling Decoder方法分析文本在多模态医学图像分割中的作用，揭示了不同数据集中文本影响的差异。
0 citationsRead paper
InstEditSeg: Instruction-Driven Image Editing for Polyp and Skin Lesion Segmentation
针对息肉和皮肤病变分割中的低对比度、模糊边界及跨域分布差异问题，提出InstEditSeg框架，通过指令驱动图像编辑并利用预训练生成模型提升分割准确性。
0 citationsRead paper
FAN-LoRA: A Fourier-Adaptive Nonlinear Low-Rank Adaptor for Medical Foundation Model Domain Adaptation
本文提出FAN-LoRA方法，通过频率解耦优化空间，解决医学图像领域适应中的性能下降问题。
0 citationsRead paper
Diverge-Merge Formation and MAC Control in Structured Airspace
In structured corridor–ramp airspace, conventional rigid formation control struggles to accommodate on-ramp merging and off-ramp splitting under high-density unmanned aerial vehicle (UAV) traffic, leading to degraded efficiency and safety. This work proposes a task-driven, real-time formation reconfiguration framework that integrates flight intent, spatial connectivity, and task-level interactions to enable dynamic geometric adaptation. Furthermore, a cluster-aware distributed TDMA protocol (CAD-TDMA) is designed to jointly optimize communication synchronization and spectrum reuse. Simulations demonstrate that the proposed approach maintains near-zero geometric misclassification even under severe congestion, while CAD-TDMA significantly outperforms fixed TDMA and WiFi MAC in terms of latency, packet loss rate, and throughput.
0 citationsRead paper
Recent publications
Latest Papers
GeoCueFormer: Geometry-Guided Wavelet Representation and Prediction-Cued Dual-Stage Decoder for Underwater Semantic Segmentation
本文提出GeoCueFormer，通过几何引导的小波表示和预测提示的双阶段解码器解决水下图像因光线吸收和散射导致的视觉退化问题，提升水下语义分割性能。
0 citationsRead paper
Characterizing Text Branch Sensitivity in Medical Vision-Language Segmentation via Evidence Decoupling
研究通过Evidence Decoupling Decoder方法分析文本在多模态医学图像分割中的作用，揭示了不同数据集中文本影响的差异。
0 citationsRead paper
InstEditSeg: Instruction-Driven Image Editing for Polyp and Skin Lesion Segmentation
针对息肉和皮肤病变分割中的低对比度、模糊边界及跨域分布差异问题，提出InstEditSeg框架，通过指令驱动图像编辑并利用预训练生成模型提升分割准确性。
0 citationsRead paper
FAN-LoRA: A Fourier-Adaptive Nonlinear Low-Rank Adaptor for Medical Foundation Model Domain Adaptation
本文提出FAN-LoRA方法，通过频率解耦优化空间，解决医学图像领域适应中的性能下降问题。
0 citationsRead paper
Diverge-Merge Formation and MAC Control in Structured Airspace
In structured corridor–ramp airspace, conventional rigid formation control struggles to accommodate on-ramp merging and off-ramp splitting under high-density unmanned aerial vehicle (UAV) traffic, leading to degraded efficiency and safety. This work proposes a task-driven, real-time formation reconfiguration framework that integrates flight intent, spatial connectivity, and task-level interactions to enable dynamic geometric adaptation. Furthermore, a cluster-aware distributed TDMA protocol (CAD-TDMA) is designed to jointly optimize communication synchronization and spectrum reuse. Simulations demonstrate that the proposed approach maintains near-zero geometric misclassification even under severe congestion, while CAD-TDMA significantly outperforms fixed TDMA and WiFi MAC in terms of latency, packet loss rate, and throughput.
0 citationsRead paper