Extending TotalSegmentator: Predicting Patient and Acquisition Characteristics from CT and MR Images
研究开发并评估了一种快速开源模型，通过CT和MR图像直接预测患者及采集特征，使用3D ResNet-10集成方法训练模型。
0 citationsRead paper
Selected work
Representative Papers
Evidence-Gated Task and Motion Planning with Vision-Language Models
本文提出EAFG框架，通过视觉证据获取和可行性判断解决机器人在执行长时操作任务中因部分可观测性导致的不确定性问题。
0 citationsRead paper
3D Underwater Path Planning via Generative Flow Field Surrogates
This study addresses the inaccuracy of conventional uniform-flow models for autonomous underwater vehicle (AUV) launch and recovery operations within propeller wake fields, where high-fidelity computational fluid dynamics (CFD) simulations are too computationally expensive for onboard real-time use. To bridge this gap, the authors propose a conditional generative adversarial network (cGAN)-based surrogate model featuring a hierarchical generative architecture and scalar operating-condition inputs, enabling end-to-end generation of 128³-voxel three-dimensional flow fields at microsecond inference speeds (28–146 µs). The synthesized flow fields are integrated into an energy-weighted A* path planner. Experimental results demonstrate that full CFD-informed planning reduces energy consumption by 5.7–12.5% and decreases traversal through high-velocity core regions by 77.8% compared to uniform-flow assumptions. The proposed cGAN recovers 45–60% of these CFD-derived benefits while remaining deployable on edge hardware, offering the first systematic quantification of the downstream planning value of generated flow fields.
0 citationsRead paper
WAKESET: A Large-Scale, High-Reynolds Number Flow Dataset for Machine Learning of Turbulent Wake Dynamics
This study addresses the scarcity of large-scale, high-fidelity datasets for machine learning in high-Reynolds-number turbulent flows. To bridge this gap, the authors construct a high-quality CFD dataset focused on turbulent wakes generated during underwater vehicle recovery, with Reynolds numbers reaching up to 1.09×10⁸ and encompassing a range of speeds and rudder angles. Leveraging high-fidelity RANS simulations, parametric case design, and data augmentation techniques, the dataset comprises 4,360 flow field instances derived from 1,091 base simulations. This work provides the first benchmark dataset of turbulent wakes tailored to real-world engineering scenarios, thereby enabling machine learning applications such as flow field prediction, surrogate modeling, and autonomous navigation in high-Reynolds-number regimes.
0 citationsRead paper
Mitigation of multi-path propagation artefacts in acoustic targets with cepstral adaptive filtering
In passive acoustic monitoring, multipath reflections and motion-induced artifacts severely degrade target signal quality; existing filtering methods lack robustness due to their neglect of environmental non-stationarity and medium heterogeneity. This paper proposes a cepstral-domain adaptive bandstop filtering method: for the first time, quefrency-intensity-driven dynamic bandwidth control is introduced for acoustic multipath suppression, enabling robust source–reflection separation in the time–frequency domain. The method integrates cepstral analysis, adaptive time–frequency filtering, and dynamic parameter adjustment to accommodate time-varying propagation conditions. Experiments on aircraft noise simulations show significant improvements in SNR, log-spectral distortion (LSD), and Itakura–Saito (IS) distance. For ship-type classification on DeepShip and VTUAD v2 datasets, Matthews correlation coefficient (MCC) increases by 2.28% and 2.62%, respectively. Moreover, target recognition accuracy and time-delay estimation precision are both notably enhanced.
0 citationsRead paper
Recent publications
Latest Papers
Extending TotalSegmentator: Predicting Patient and Acquisition Characteristics from CT and MR Images
研究开发并评估了一种快速开源模型，通过CT和MR图像直接预测患者及采集特征，使用3D ResNet-10集成方法训练模型。
0 citationsRead paper
Evidence-Gated Task and Motion Planning with Vision-Language Models
本文提出EAFG框架，通过视觉证据获取和可行性判断解决机器人在执行长时操作任务中因部分可观测性导致的不确定性问题。
0 citationsRead paper
3D Underwater Path Planning via Generative Flow Field Surrogates
This study addresses the inaccuracy of conventional uniform-flow models for autonomous underwater vehicle (AUV) launch and recovery operations within propeller wake fields, where high-fidelity computational fluid dynamics (CFD) simulations are too computationally expensive for onboard real-time use. To bridge this gap, the authors propose a conditional generative adversarial network (cGAN)-based surrogate model featuring a hierarchical generative architecture and scalar operating-condition inputs, enabling end-to-end generation of 128³-voxel three-dimensional flow fields at microsecond inference speeds (28–146 µs). The synthesized flow fields are integrated into an energy-weighted A* path planner. Experimental results demonstrate that full CFD-informed planning reduces energy consumption by 5.7–12.5% and decreases traversal through high-velocity core regions by 77.8% compared to uniform-flow assumptions. The proposed cGAN recovers 45–60% of these CFD-derived benefits while remaining deployable on edge hardware, offering the first systematic quantification of the downstream planning value of generated flow fields.
0 citationsRead paper
WAKESET: A Large-Scale, High-Reynolds Number Flow Dataset for Machine Learning of Turbulent Wake Dynamics
This study addresses the scarcity of large-scale, high-fidelity datasets for machine learning in high-Reynolds-number turbulent flows. To bridge this gap, the authors construct a high-quality CFD dataset focused on turbulent wakes generated during underwater vehicle recovery, with Reynolds numbers reaching up to 1.09×10⁸ and encompassing a range of speeds and rudder angles. Leveraging high-fidelity RANS simulations, parametric case design, and data augmentation techniques, the dataset comprises 4,360 flow field instances derived from 1,091 base simulations. This work provides the first benchmark dataset of turbulent wakes tailored to real-world engineering scenarios, thereby enabling machine learning applications such as flow field prediction, surrogate modeling, and autonomous navigation in high-Reynolds-number regimes.
0 citationsRead paper
Mitigation of multi-path propagation artefacts in acoustic targets with cepstral adaptive filtering
In passive acoustic monitoring, multipath reflections and motion-induced artifacts severely degrade target signal quality; existing filtering methods lack robustness due to their neglect of environmental non-stationarity and medium heterogeneity. This paper proposes a cepstral-domain adaptive bandstop filtering method: for the first time, quefrency-intensity-driven dynamic bandwidth control is introduced for acoustic multipath suppression, enabling robust source–reflection separation in the time–frequency domain. The method integrates cepstral analysis, adaptive time–frequency filtering, and dynamic parameter adjustment to accommodate time-varying propagation conditions. Experiments on aircraft noise simulations show significant improvements in SNR, log-spectral distortion (LSD), and Itakura–Saito (IS) distance. For ship-type classification on DeepShip and VTUAD v2 datasets, Matthews correlation coefficient (MCC) increases by 2.28% and 2.62%, respectively. Moreover, target recognition accuracy and time-delay estimation precision are both notably enhanced.
0 citationsRead paper