Wai Keung Wong

Dynamic hand gesture authentication (DHGA) enables touchless identity verification from gesture video by jointly leveraging physiological cues and behavioral dynamics, offering inherent spoofing resistance. However, existing methods focus heavily on behavioral dynamics while largely overlooking the physiologically discriminative appearance cues, leaving authentication vulnerable to environmental perturbation. In this paper, we propose an edge-anchored spatio-temporal learning network (EAST-Net) that explicitly decomposes physiological identity information into complementary appearance and structural representations, while jointly modeling their temporal evolution to capture behavioral identity cues through a three-stage pipeline. We first explicitly decompose spatial identity information into RGB-based appearance features and Canny-based structural edge cues via parallel streams, each embedded with a micro temporal shift module to capture subtle behavioral dynamics while preserving spatial fidelity. Then, we align the appearance and structural features via spatial cross-attention, anchoring fragile appearance details onto robust structural topologies to resist visual degradation. Finally, we apply a squeeze-and-excitation fusion module to adaptively consolidate cross-modal features for DHGA, balancing discriminative power with environmental robustness. Extensive experiments show that EAST-Net consistently outperforms state-of-the-art methods.

Recent advances in multi-view fundus imaging show great promise for automated diabetic retinopathy (DR) grading. However, mainstream end-to-end CNN/Transformer pipelines rely on striding or tokenization that compresses spatial detail, causing small, low-contrast lesions (e.g., microaneurysms) to be under-represented and creating performance ceilings. Prior efforts have mitigated this by incorporating external lesion- or vessel-level annotations into models. However, such labels are costly to acquire, break the end-to-end training, and make performance over-reliant on the annotation quality. To reduce dependence on expensive annotations, we propose an end-to-end framework that generates lesion proposals on the fly during training and inference, providing self-derived cues for grading. First, we introduce a Grade-Activated Lesion Proposal (GALP) module that derives grade-conditioned evidence maps (GEMs) from stage-wise auxiliary classifiers and selects the top-K high-evidence regions per view as lesion proposals. Second, we propose a Cross-View Lesion Expert Guided Regional Fusion (LGRF) module, which selectively activates experts for a view’s lesion proposals based on contextual guidance from other views, ensuring that only the most relevant feature extractors contribute to fusion. Experimental results on two multi-view DR datasets show that our method matches or surpasses strong baselines without external annotations, confirming that self-generated proposals can substantially reduce annotation needs.

Timely identification of Alzheimer’s disease (AD) benefits from combining neuroimaging, fluid biomarkers, and cognitive assessments, yet in practice one or more modalities are often unavailable due to various factors such as cost, patient compliance, and procedural risks. Furthermore, conventional convolutional neural network (CNN) architectures and even Transformer-based models struggle to efficiently capture both local and global dependencies, especially when dealing with high-dimensional and highly heterogeneous medical data. In this study, we introduce a novel hierarchical information aggregation and dynamic fusion (HI-AD) framework for incomplete multi-modal AD diagnosis. Our method couples a multi-view Hilbert curve–guided Mamba block with hierarchical spatial feature extraction to retain spatial continuity, model long-range dependencies, and integrate local context in neuroimaging data. To balance semantic alignment and modality-specific information, we propose a unified mutual information-driven learning objective with an active confidence evaluation mechanism, thereby preventing modality collapse and promoting robust representation learning. Extensive experiments on real-world datasets validate that our HI-AD framework consistently outperforms existing state-of-the-art methods across a diverse range of modality-missing scenarios, establishing an effective and generalizable solution for early-stage AD screening in heterogeneous clinical data environments.

In recent years, numerous hashing techniques have been developed to boost efficient cross-modal retrieval. Once a retrieval model is deployed, the hash code length is fixed to achieve optimal performance. To address different retrieval scenarios while maintaining retrieval accuracy, a common approach is to redesign and retrain the original model with different hash code length. However, this retraining process can increase the training load and may lead to worse results. To tackle these challenges, we present Regenerated Cross-Modal Hashing (RCMH), a novel cross-modal hashing framework designed to improve the quality of existing hash codes and convert them to arbitrary lengths with high efficiency. First, we clip or pad the existing hash codes to initialize them with the target length, under the supervision of the similarity matrix generated by the augmented label information. Second, we introduce a linear-nonlinear competitive reconstruction approach to reduce the semantic gaps and further capture the deeper relationships from linear image features and nonlinear text features. In this way, each pair of samples is compared and selected to obtain reconstructed binary codes that can preserve the modality-specific properties. Finally, to reduce the training costs caused by iterations of variables, the regenerate hashing term is utilized to regenerate final hash codes with the reconstructed binary codes while preserving the information from the existing hash codes without iterative optimization. Notably, RCMH can be integrated with existing state-of-the-art (SOTA) methods with robustness, helping them to adjust the hash code length and achieve better retrieval performance. Extensive experiments on several public datasets show that RCMH outperforms SOTA methods across various evaluation metrics. The code is available at https://github.com/kaihang-jiang/RCMH.git.

Deep multi-view clustering aims to exploit the rich semantic information contained in heterogeneous multi-view data to uncover the underlying relationships among samples. However, existing deep multi-view clustering models often overlook intercluster separability and the effective integration of semantic information across views, resulting in insufficient feature discriminability and consequently limited clustering performance. To address the above issues, this paper proposes a novel deep multi-view clustering method via cluster-semantic guidance. We separate clusters to enhance inter-cluster discriminability, while incorporating a knowledge distillation mechanism to ensure cluster stability and facilitate the learning of clustering-friendly representations. Furthermore, by aggregating sample-level semantic information, the model is guided to follow a cluster-oriented learning strategy that promotes the extraction of discriminative features, thereby strengthening the sample representation capability. Our method effectively learns discriminative and clustering-friendly representations, guiding the model to acquire distinctive feature embeddings from a cluster-oriented perspective. Our comprehensive experiments across datasets of varying scales confirm the model’s effectiveness, showing superior clustering performance over existing state-of-the-art methods.

Image-based Virtual Try-On (VTON) aims to generate virtual try-on results by transferring input garment images onto target person images. Current algorithms have achieved high generalization and quality on public datasets including VITON and DressCode. However, current algorithms primarily process images at the pixel level while neglecting the semantic information of garments. Moreover, layered outerwear, as an essential aspect of try-on, has been largely overlooked. Layering increases the difficulty of precise garment control due to the need to consider inner garments. Furthermore, the task-specific nature of current methods prevents them from capturing layering characteristics in try-on. To address these limitations, we propose LayerDiffusion, a novel framework that enhances both generalization capability and semantic controllability for virtual try-on generation. Our approach makes three key contributions. First, we construct LayerDataset, a specialized dataset focusing on diverse outerwear garments with balanced gender representation and multi-view captures. Second, we integrate a pretrained image encoder to capture rich semantic garment information, combined with multi-conditional masking and adapter fine-tuning to enable flexible layering control. Third, we introduce a joint training strategy on both try-on and try-off tasks, which substantially improves model generalization by learning bidirectional garment-body correlations. Extensive experiments demonstrate that LayerDiffusion significantly outperforms existing state-of-the-art methods on public benchmarks (VITON-HD and DressCode) across multiple metrics including LPIPS, SSIM, FID, KID, and CLIP-based scores. Our method excels in both single-garment and layered try-on scenarios while naturally supporting try-off capabilities, offering superior semantic preservation and detail fidelity compared to existing approaches.