Kun Hu

Accurate and cost-effective estimation of soil organic carbon (SOC) is essential for monitoring soil health, supporting sustainable land management, and improving soil carbon accounting. Visible and near-infrared (VIS-NIR) spectroscopy offers a rapid, non-destructive alternative to conventional laboratory analysis, but reliable SOC prediction requires representative spectral libraries with corresponding analytical measurements. Prediction is further challenged by complex soil composition and heterogeneous sample conditions. In this study, we propose a Self-Supervised Multi-View Neural Network (S^2MVNN) for SOC estimation from VIS-NIR soil spectra using unlabelled data. The framework represents each spectrum through two complementary views: the original wavelength-domain spectrum and a spectrogram representation that captures local scale-dependent spectral variation. A masked reconstruction-based self-supervised strategy is used to learn spectral representations from large volumes of unlabelled spectra, reducing reliance on labelled SOC measurements. The learned spectral and spectrogram features are then fused for supervised SOC prediction. Experiments were conducted using a large Australian soil spectral library. S^2MVNN outperformed conventional baselines, including partial least squares regression, gradient boosting, and multilayer perceptron models. On the held-out test set, the model achieved an RPD of 1.914 and Lin's concordance correlation coefficient of 0.859, indicating improved agreement with laboratory-measured SOC values. These results suggest that self-supervised multi-view learning can enhance the use of large soil spectral libraries for scalable SOC estimation, while further validation across independent datasets and acquisition conditions remains important.

Recognizing human actions from point cloud sequences is critical for 3D perception driven applications such as autonomous driving and human-computer interaction. However, the irregular structure and temporal inconsistency of point clouds pose unique challenges for spatio-temporal representation learning, especially in capturing both global motion context and fine-grained temporal dynamics. We propose SRENet, a spectral-aware framework designed to explicitly learn both global context and fine-grained temporal dynamics of motion from a frequency perspective for action recognition. SRENet introduces a Spectral Decomposition Block (SDeBlock) that performs wavelet-based analysis along temporal and spatial axes, disentangling features into low- and high-frequency components with frequency-specific attention. To recover residual dynamics and re-align temporal frequency structures distorted during semantic fusion, a Spectral Re-entry Block (SReBlock) performs secondary temporal decomposition. Furthermore, a spectral-aware learning strategy is devised to enhance discriminability in both frequency subspaces via contrastive loss and a curriculum schedule that gradually shifts focus from low- to high-frequency spaces in line with coarse to detailed motion patterns. Extensive experiments on MSR-Action3D, NTU-RGBD and NTU-RGBD120 demonstrate that SRENet achieves state-of-the-art performance, validating the effectiveness of frequency modeling in point cloud-based action understanding. ¹.

Garment simulation is fundamental to various applications in computer vision and graphics, from virtual try-on to digital human modelling. However, conventional physics-based methods remain computationally expensive, hindering their application in time-sensitive scenarios. While graph neural networks (GNNs) offer promising acceleration, existing approaches exhibit poor cross-resolution generalisation, demonstrating significant performance degradation on higher-resolution meshes beyond the training distribution. This stems from two key factors: (1) existing GNNs employ fixed message-passing depth that fails to adapt information aggregation to mesh density variation, and (2) vertex-wise displacement magnitudes are inherently resolution-dependent in garment simulation. To address these issues, we introduce Propagation-before-Update Graph Network (Pb4U-GNet), a resolution-adaptive framework that decouples message propagation from feature updates. Pb4U-GNet incorporates two key mechanisms: (1) dynamic propagation depth control, adjusting message-passing iterations based on mesh resolution, and (2) geometry-aware update scaling, which scales predictions according to local mesh characteristics. Extensive experiments show that even trained solely on low-resolution meshes, Pb4U-GNet exhibits strong generalisability across diverse mesh resolutions, addressing a fundamental challenge in neural garment simulation.