arXiv subjects

Zhihui Lai

Publications and source records attributed to Zhihui Lai.

At least 19 recordsLinked to original sources

arXiv2026

FGTBT: Frequency-Guided Task-Balancing Transformer for Unified Facial Landmark Detection

Recently, deep learning based facial landmark detection (FLD) methods have achieved considerable success. However, in challenging scenarios such as large pose variations, illumination changes, and facial expression variations, they still struggle to accurately capture the geometric structure of the face, resulting in performance degradation. Moreover, the limited size and diversity of existing FLD datasets hinder robust model training, leading to reduced detection accuracy. To address these challenges, we propose a Frequency-Guided Task-Balancing Transformer (FGTBT), which enhances facial structure perception through frequency-domain modeling and multi-dataset unified training. Specifically, we propose a novel Fine-Grained Multi-Task Balancing loss (FMB-loss), which moves beyond coarse task-level balancing by assigning weights to individual landmarks based on their occurrence across datasets. This enables more effective unified training and mitigates the issue of inconsistent gradient magnitudes. Additionally, a Frequency-Guided Structure-Aware (FGSA) model is designed to utilize frequency-guided structure injection and regularization to help learn facial structure constraints. Extensive experimental results on popular benchmark datasets demonstrate that the integration of the proposed FMB-loss and FGSA model into our FGTBT framework achieves performance comparable to state-of-the-art methods. The code is available at https://github.com/Xi0ngxinyu/FGTBT.

cs.CV↗

arXiv2026

Supervision-by-Hallucination-and-Transfer: A Weakly-Supervised Approach for Robust and Precise Facial Landmark Detection

High-precision facial landmark detection (FLD) relies on high-resolution deep feature representations. However, low-resolution face images or the compression (via pooling or strided convolution) of originally high-resolution images hinder the learning of such features, thereby reducing FLD accuracy. Moreover, insufficient training data and imprecise annotations further degrade performance. To address these challenges, we propose a weakly-supervised framework called Supervision-by-Hallucination-and-Transfer (SHT) for more robust and precise FLD. SHT contains two novel mutually enhanced modules: Dual Hallucination Learning Network (DHLN) and Facial Pose Transfer Network (FPTN). By incorporating FLD and face hallucination tasks, DHLN is able to learn high-resolution representations with low-resolution inputs for recovering both facial structures and local details and generating more effective landmark heatmaps. Then, by transforming faces from one pose to another, FPTN can further improve landmark heatmaps and faces hallucinated by DHLN for detecting more accurate landmarks. To the best of our knowledge, this is the first study to explore weakly-supervised FLD by integrating face hallucination and facial pose transfer tasks. Experimental results of both face hallucination and FLD demonstrate that our method surpasses state-of-the-art techniques.

cs.CV↗

arXiv2025

FaNe: Towards Fine-Grained Cross-Modal Contrast with False-Negative Reduction and Text-Conditioned Sparse Attention

Medical vision-language pre-training (VLP) offers significant potential for advancing medical image understanding by leveraging paired image-report data. However, existing methods are limited by Fa}lse Negatives (FaNe) induced by semantically similar texts and insufficient fine-grained cross-modal alignment. To address these limitations, we propose FaNe, a semantic-enhanced VLP framework. To mitigate false negatives, we introduce a semantic-aware positive pair mining strategy based on text-text similarity with adaptive normalization. Furthermore, we design a text-conditioned sparse attention pooling module to enable fine-grained image-text alignment through localized visual representations guided by textual cues. To strengthen intra-modal discrimination, we develop a hard-negative aware contrastive loss that adaptively reweights semantically similar negatives. Extensive experiments on five downstream medical imaging benchmarks demonstrate that FaNe achieves state-of-the-art performance across image classification, object detection, and semantic segmentation, validating the effectiveness of our framework.

cs.CV↗

arXiv2025

LightQANet: Quantized and Adaptive Feature Learning for Low-Light Image Enhancement

Low-light image enhancement (LLIE) aims to improve illumination while preserving high-quality color and texture. However, existing methods often fail to extract reliable feature representations due to severely degraded pixel-level information under low-light conditions, resulting in poor texture restoration, color inconsistency, and artifact. To address these challenges, we propose LightQANet, a novel framework that introduces quantized and adaptive feature learning for low-light enhancement, aiming to achieve consistent and robust image quality across diverse lighting conditions. From the static modeling perspective, we design a Light Quantization Module (LQM) to explicitly extract and quantify illumination-related factors from image features. By enforcing structured light factor learning, LQM enhances the extraction of light-invariant representations and mitigates feature inconsistency across varying illumination levels. From the dynamic adaptation perspective, we introduce a Light-Aware Prompt Module (LAPM), which encodes illumination priors into learnable prompts to dynamically guide the feature learning process. LAPM enables the model to flexibly adapt to complex and continuously changing lighting conditions, further improving image enhancement. Extensive experiments on multiple low-light datasets demonstrate that our method achieves state-of-the-art performance, delivering superior qualitative and quantitative results across various challenging lighting scenarios.

cs.CV↗

arXiv2025

Semi-Supervised Dual-Threshold Contrastive Learning for Ultrasound Image Classification and Segmentation

Confidence-based pseudo-label selection usually generates overly confident yet incorrect predictions, due to the early misleadingness of model and overfitting inaccurate pseudo-labels in the learning process, which heavily degrades the performance of semi-supervised contrastive learning. Moreover, segmentation and classification tasks are treated independently and the affinity fails to be fully explored. To address these issues, we propose a novel semi-supervised dual-threshold contrastive learning strategy for ultrasound image classification and segmentation, named Hermes. This strategy combines the strengths of contrastive learning with semi-supervised learning, where the pseudo-labels assist contrastive learning by providing additional guidance. Specifically, an inter-task attention and saliency module is also developed to facilitate information sharing between the segmentation and classification tasks. Furthermore, an inter-task consistency learning strategy is designed to align tumor features across both tasks, avoiding negative transfer for reducing features discrepancy. To solve the lack of publicly available ultrasound datasets, we have collected the SZ-TUS dataset, a thyroid ultrasound image dataset. Extensive experiments on two public ultrasound datasets and one private dataset demonstrate that Hermes consistently outperforms several state-of-the-art methods across various semi-supervised settings.

cs.CV↗

arXiv2025

CodeEnhance: A Codebook-Driven Approach for Low-Light Image Enhancement

Low-light image enhancement (LLIE) aims to improve low-illumination images. However, existing methods face two challenges: (1) uncertainty in restoration from diverse brightness degradations; (2) loss of texture and color information caused by noise suppression and light enhancement. In this paper, we propose a novel enhancement approach, CodeEnhance, by leveraging quantized priors and image refinement to address these challenges. In particular, we reframe LLIE as learning an image-to-code mapping from low-light images to discrete codebook, which has been learned from high-quality images. To enhance this process, a Semantic Embedding Module (SEM) is introduced to integrate semantic information with low-level features, and a Codebook Shift (CS) mechanism, designed to adapt the pre-learned codebook to better suit the distinct characteristics of our low-light dataset. Additionally, we present an Interactive Feature Transformation (IFT) module to refine texture and color information during image reconstruction, allowing for interactive enhancement based on user preferences. Extensive experiments on both real-world and synthetic benchmarks demonstrate that the incorporation of prior knowledge and controllable information transfer significantly enhances LLIE performance in terms of quality and fidelity. The proposed CodeEnhance exhibits superior robustness to various degradations, including uneven illumination, noise, and color distortion.

cs.CV↗

arXiv2025

Joker: Joint Optimization Framework for Lightweight Kernel Machines

Kernel methods are powerful tools for nonlinear learning with well-established theory. The scalability issue has been their long-standing challenge. Despite the existing success, there are two limitations in large-scale kernel methods: (i) The memory overhead is too high for users to afford; (ii) existing efforts mainly focus on kernel ridge regression (KRR), while other models lack study. In this paper, we propose Joker, a joint optimization framework for diverse kernel models, including KRR, logistic regression, and support vector machines. We design a dual block coordinate descent method with trust region (DBCD-TR) and adopt kernel approximation with randomized features, leading to low memory costs and high efficiency in large-scale learning. Experiments show that Joker saves up to 90\% memory but achieves comparable training time and performance (or even better) than the state-of-the-art methods.

cs.LG↗

arXiv2025

Prototype Contrastive Consistency Learning for Semi-Supervised Medical Image Segmentation

Medical image segmentation is a crucial task in medical image analysis, but it can be very challenging especially when there are less labeled data but with large unlabeled data. Contrastive learning has proven to be effective for medical image segmentation in semi-supervised learning by constructing contrastive samples from partial pixels. However, although previous contrastive learning methods can mine semantic information from partial pixels within images, they ignore the whole context information of unlabeled images, which is very important to precise segmentation. In order to solve this problem, we propose a novel prototype contrastive learning method called Prototype Contrastive Consistency Segmentation (PCCS) for semi-supervised medical image segmentation. The core idea is to enforce the prototypes of the same semantic class to be closer and push the prototypes in different semantic classes far away from each other. Specifically, we construct a signed distance map and an uncertainty map from unlabeled images. The signed distance map is used to construct prototypes for contrastive learning, and then we estimate the prototype uncertainty from the uncertainty map as trade-off among prototypes. In order to obtain better prototypes, based on the student-teacher architecture, a new mechanism named prototype updating prototype is designed to assist in updating the prototypes for contrastive learning. In addition, we propose an uncertainty-consistency loss to mine more reliable information from unlabeled data. Extensive experiments on medical image segmentation demonstrate that PCCS achieves better segmentation performance than the state-of-the-art methods. The code is available at https://github.com/comphsh/PCCS.

cs.CV↗

arXiv2024

Precise Facial Landmark Detection by Dynamic Semantic Aggregation Transformer

At present, deep neural network methods have played a dominant role in face alignment field. However, they generally use predefined network structures to predict landmarks, which tends to learn general features and leads to mediocre performance, e.g., they perform well on neutral samples but struggle with faces exhibiting large poses or occlusions. Moreover, they cannot effectively deal with semantic gaps and ambiguities among features at different scales, which may hinder them from learning efficient features. To address the above issues, in this paper, we propose a Dynamic Semantic-Aggregation Transformer (DSAT) for more discriminative and representative feature (i.e., specialized feature) learning. Specifically, a Dynamic Semantic-Aware (DSA) model is first proposed to partition samples into subsets and activate the specific pathways for them by estimating the semantic correlations of feature channels, making it possible to learn specialized features from each subset. Then, a novel Dynamic Semantic Specialization (DSS) model is designed to mine the homogeneous information from features at different scales for eliminating the semantic gap and ambiguities and enhancing the representation ability. Finally, by integrating the DSA model and DSS model into our proposed DSAT in both dynamic architecture and dynamic parameter manners, more specialized features can be learned for achieving more precise face alignment. It is interesting to show that harder samples can be handled by activating more feature channels. Extensive experiments on popular face alignment datasets demonstrate that our proposed DSAT outperforms state-of-the-art models in the literature.Our code is available at https://github.com/GERMINO-LiuHe/DSAT.

cs.CV↗

arXiv2024

Low-Light Enhancement Effect on Classification and Detection: An Empirical Study

Low-light images are commonly encountered in real-world scenarios, and numerous low-light image enhancement (LLIE) methods have been proposed to improve the visibility of these images. The primary goal of LLIE is to generate clearer images that are more visually pleasing to humans. However, the impact of LLIE methods in high-level vision tasks, such as image classification and object detection, which rely on high-quality image datasets, is not well {explored}. To explore the impact, we comprehensively evaluate LLIE methods on these high-level vision tasks by utilizing an empirical investigation comprising image classification and object detection experiments. The evaluation reveals a dichotomy: {\textit{While Low-Light Image Enhancement (LLIE) methods enhance human visual interpretation, their effect on computer vision tasks is inconsistent and can sometimes be harmful. }} Our findings suggest a disconnect between image enhancement for human visual perception and for machine analysis, indicating a need for LLIE methods tailored to support high-level vision tasks effectively. This insight is crucial for the development of LLIE techniques that align with the needs of both human and machine vision.

cs.CV↗

arXiv2024

NPSVC++: Nonparallel Classifiers Encounter Representation Learning

This paper focuses on a specific family of classifiers called nonparallel support vector classifiers (NPSVCs). Different from typical classifiers, the training of an NPSVC involves the minimization of multiple objectives, resulting in the potential concerns of feature suboptimality and class dependency. Consequently, no effective learning scheme has been established to improve NPSVCs' performance through representation learning, especially deep learning. To break this bottleneck, we develop NPSVC++ based on multi-objective optimization, enabling the end-to-end learning of NPSVC and its features. By pursuing Pareto optimality, NPSVC++ theoretically ensures feature optimality across classes, hence effectively overcoming the two issues above. A general learning procedure via duality optimization is proposed, based on which we provide two applicable instances, K-NPSVC++ and D-NPSVC++. The experiments show their superiority over the existing methods and verify the efficacy of NPSVC++.

cs.LG↗

arXiv2024

A Nonlinear Damped Metamaterial: Wideband Attenuation with Nonlinear Bandgap and Modal Dissipation

In this paper, we incorporate the effect of nonlinear damping with the concept of locally resonant metamaterials to enable vibration attenuation beyond the conventional bandgap range. The proposed design combines a linear host cantilever beam and periodically distributed inertia amplifiers as nonlinear local resonators. The geometric nonlinearity induced by the inertia amplifiers causes an amplitude-dependent nonlinear damping effect. Through the implementation of both modal superposition and numerical harmonic methods the finite nonlinear metamaterial is accurately modelled. The resulting nonlinear frequency response reveals the bandgap is both amplitude-dependent and broadened. Furthermore, the modal frequencies are also attenuated due to the nonlinear damping effect. The theoretical results are validated experimentally. By embedding the nonlinear damping effect into locally resonant metamaterials, wideband attenuation of the proposed metamaterial is achieved, which opens new possibilities for versatile metamaterials beyond the limit of their linear counterparts.

physics.app-ph↗

arXiv2023

Computational Technologies for Fashion Recommendation: A Survey

Fashion recommendation is a key research field in computational fashion research and has attracted considerable interest in the computer vision, multimedia, and information retrieval communities in recent years. Due to the great demand for applications, various fashion recommendation tasks, such as personalized fashion product recommendation, complementary (mix-and-match) recommendation, and outfit recommendation, have been posed and explored in the literature. The continuing research attention and advances impel us to look back and in-depth into the field for a better understanding. In this paper, we comprehensively review recent research efforts on fashion recommendation from a technological perspective. We first introduce fashion recommendation at a macro level and analyse its characteristics and differences with general recommendation tasks. We then clearly categorize different fashion recommendation efforts into several sub-tasks and focus on each sub-task in terms of its problem formulation, research focus, state-of-the-art methods, and limitations. We also summarize the datasets proposed in the literature for use in fashion recommendation studies to give readers a brief illustration. Finally, we discuss several promising directions for future research in this field. Overall, this survey systematically reviews the development of fashion recommendation research. It also discusses the current limitations and gaps between academic research and the real needs of the fashion industry. In the process, we offer a deep insight into how the fashion industry could benefit from the computational technologies of fashion recommendation.

cs.MM↗

arXiv2023

Precise Facial Landmark Detection by Reference Heatmap Transformer

Most facial landmark detection methods predict landmarks by mapping the input facial appearance features to landmark heatmaps and have achieved promising results. However, when the face image is suffering from large poses, heavy occlusions and complicated illuminations, they cannot learn discriminative feature representations and effective facial shape constraints, nor can they accurately predict the value of each element in the landmark heatmap, limiting their detection accuracy. To address this problem, we propose a novel Reference Heatmap Transformer (RHT) by introducing reference heatmap information for more precise facial landmark detection. The proposed RHT consists of a Soft Transformation Module (STM) and a Hard Transformation Module (HTM), which can cooperate with each other to encourage the accurate transformation of the reference heatmap information and facial shape constraints. Then, a Multi-Scale Feature Fusion Module (MSFFM) is proposed to fuse the transformed heatmap features and the semantic features learned from the original face images to enhance feature representations for producing more accurate target heatmaps. To the best of our knowledge, this is the first study to explore how to enhance facial landmark detection by transforming the reference heatmap information. The experimental results from challenging benchmark datasets demonstrate that our proposed method outperforms the state-of-the-art methods in the literature.

cs.CV↗

arXiv2023

Harmonic-Gaussian double-well potential stochastic resonance with its application to enhance weak fault characteristics of machinery

Noise is ubiquitous and unwanted in detecting weak signals, which would give rise to incorrect filtering frequency-band selection in signal filtering-based methods including fast kurtogram, teager energy operators and wavelet packet transform filters and meanwhile would result in incorrect selection of useful components and even mode mixing, end effects and etc. in signal decomposition-based methods including empirical mode decomposition, singular value decomposition and local mean decomposition. On the contrary, noise in stochastic resonance (SR) is beneficial to enhance weak signals of interest embedded in signals with strong background noise. Taking into account that nonlinear systems are crucial ingredients to activate the SR, here we investigate the SR in the cases of overdamped and underdamped harmonic-Gaussian double-well potential systems subjected to noise and a periodic signal. We derive and measure the analytic expression of the output signal-to-noise ratio (SNR) and the steady-state probability density (SPD) function under approximate adiabatic conditions. When the harmonic-Gaussian double-well potential loses its stability, we can observe the antiresonance phenomenon, whereas adding the damped factor into the overdamped system can change the stability of the harmonic-Gaussian double-well potential, resulting that the antiresonance behavior disappears in the underdamped system. Then, we use the overdamped and underdamped harmonic-Gaussian double-well potential SR to enhance weak useful characteristics for diagnosing incipient rotating machinery failures.

physics.app-ph↗

arXiv2021

Robust and Precise Facial Landmark Detection by Self-Calibrated Pose Attention Network

Current fully-supervised facial landmark detection methods have progressed rapidly and achieved remarkable performance. However, they still suffer when coping with faces under large poses and heavy occlusions for inaccurate facial shape constraints and insufficient labeled training samples. In this paper, we propose a semi-supervised framework, i.e., a Self-Calibrated Pose Attention Network (SCPAN) to achieve more robust and precise facial landmark detection in challenging scenarios. To be specific, a Boundary-Aware Landmark Intensity (BALI) field is proposed to model more effective facial shape constraints by fusing boundary and landmark intensity field information. Moreover, a Self-Calibrated Pose Attention (SCPA) model is designed to provide a self-learned objective function that enforces intermediate supervision without label information by introducing a self-calibrated mechanism and a pose attention mask. We show that by integrating the BALI fields and SCPA model into a novel self-calibrated pose attention network, more facial prior knowledge can be learned and the detection accuracy and robustness of our method for faces with large poses and heavy occlusions have been improved. The experimental results obtained for challenging benchmark datasets demonstrate that our approach outperforms state-of-the-art methods in the literature.

cs.CV↗

arXiv2021

Multi-Tensor Network Representation for High-Order Tensor Completion

This work studies the problem of high-dimensional data (referred to as tensors) completion from partially observed samplings. We consider that a tensor is a superposition of multiple low-rank components. In particular, each component can be represented as multilinear connections over several latent factors and naturally mapped to a specific tensor network (TN) topology. In this paper, we propose a fundamental tensor decomposition (TD) framework: Multi-Tensor Network Representation (MTNR), which can be regarded as a linear combination of a range of TD models, e.g., CANDECOMP/PARAFAC (CP) decomposition, Tensor Train (TT), and Tensor Ring (TR). Specifically, MTNR represents a high-order tensor as the addition of multiple TN models, and the topology of each TN is automatically generated instead of manually pre-designed. For the optimization phase, an adaptive topology learning (ATL) algorithm is presented to obtain latent factors of each TN based on a rank incremental strategy and a projection error measurement strategy. In addition, we theoretically establish the fundamental multilinear operations for the tensors with TN representation, and reveal the structural transformation of MTNR to a single TN. Finally, MTNR is applied to a typical task, tensor completion, and two effective algorithms are proposed for the exact recovery of incomplete data based on the Alternating Least Squares (ALS) scheme and Alternating Direction Method of Multiplier (ADMM) framework. Extensive numerical experiments on synthetic data and real-world datasets demonstrate the effectiveness of MTNR compared with the start-of-the-art methods.

cs.CV↗

arXiv2021

WaveCNet: Wavelet Integrated CNNs to Suppress Aliasing Effect for Noise-Robust Image Classification

Though widely used in image classification, convolutional neural networks (CNNs) are prone to noise interruptions, i.e. the CNN output can be drastically changed by small image noise. To improve the noise robustness, we try to integrate CNNs with wavelet by replacing the common down-sampling (max-pooling, strided-convolution, and average pooling) with discrete wavelet transform (DWT). We firstly propose general DWT and inverse DWT (IDWT) layers applicable to various orthogonal and biorthogonal discrete wavelets like Haar, Daubechies, and Cohen, etc., and then design wavelet integrated CNNs (WaveCNets) by integrating DWT into the commonly used CNNs (VGG, ResNets, and DenseNet). During the down-sampling, WaveCNets apply DWT to decompose the feature maps into the low-frequency and high-frequency components. Containing the main information including the basic object structures, the low-frequency component is transmitted into the following layers to generate robust high-level features. The high-frequency components are dropped to remove most of the data noises. The experimental results show that %wavelet accelerates the CNN training, and WaveCNets achieve higher accuracy on ImageNet than various vanilla CNNs. We have also tested the performance of WaveCNets on the noisy version of ImageNet, ImageNet-C and six adversarial attacks, the results suggest that the proposed DWT/IDWT layers could provide better noise-robustness and adversarial robustness. When applying WaveCNets as backbones, the performance of object detectors (i.e., faster R-CNN and RetinaNet) on COCO detection dataset are consistently improved. We believe that suppression of aliasing effect, i.e. separation of low frequency and high frequency information, is the main advantages of our approach. The code of our DWT/IDWT layer and different WaveCNets are available at https://github.com/CVI-SZU/WaveCNet.

cs.CV↗