arXiv subjects

Pengfei Zhu

Publications and source records attributed to Pengfei Zhu.

At least 73 records · Page 4Linked to original sources

arXiv2025

Real-Time Super-Resolution Imaging System Based on Zero-Shot Learning for Infrared Non-Destructive Testing

Infrared thermography (IRT) and photothermal coherence tomography (PCT) exhibit potential in non-destructive testing and biomedical fields. However, the inevitable heat diffusion significantly affects the sensitivity and resolution of IRT and PCT. Conventional image processing techniques rely on capturing complete thermal sequences, which limits their ability to achieve real-time processes. Here, we construct a real-time super-resolution imaging system based on zero-shot learning strategy for the non-invasive infrared thermography and photothermal coherence tomography techniques. To validate the feasibility and accuracy of this super-resolution imaging system, IRT systems were employed to test several industrial samples and one biomedical sample. The results demonstrated high contrast in the region of interest (ROI) and uncovered valuable information otherwise obscured by thermal diffusion. Furthermore, three-dimensional photothermal coherence tomography was used to validate the excellent denoising and deconvolution capabilities of the proposed real-time super-resolution imaging system.

physics.app-ph↗

arXiv2025

Making neural networks understand internal heat transfer using Fourier-transformed thermal diffusion wave fields

Heat propagation is governed by phonon interactions and mathematically described by partial differential equations (PDEs), which link thermal transport to the intrinsic properties of materials. Conventional experimental techniques infer thermal responses based on surface emissions, limiting their ability to fully resolve subsurface structures and internal heat distribution. Additionally, existing thermal tomographic techniques can only shoot one frame from each layer. Physics-informed neural networks (PINNs) have recently emerged as powerful tools for solving inverse problems in heat transfer by integrating observational data with physical constraints. However, standard PINNs are primarily focused on fitting the given external temperature data, without explicit knowledge of the unknown internal temperature distribution. In this study, we introduce a Helmholtz-informed neural network (HINN) to predict internal temperature distributions without requiring internal measurements. The time-domain heat diffusion equation was converted to the frequency-domain and becomes the pseudo-Helmholtz equation. HINN embeds this pseudo-Helmholtz equation into the learning framework, leveraging both real and imaginary components of the thermal field. Finally, an inverse Fourier transform brings real-part and imagery-part back to the time-domain and can be used to map 3D thermal fields with interior defects. Furthermore, a truncated operation was conducted to improve computational efficiency, and the principle of conjugate symmetry was employed for repairing the discarded data. This approach significantly enhances predictive accuracy and computational efficiency. Our results demonstrate that HINN outperforms state-of-the-art PINNs and inverse heat solvers, offering a novel solution for non-invasive thermography in applications spanning materials science, biomedical diagnostics, and nondestructive evaluation.

physics.app-ph↗

arXiv2025

StableIntrinsic: Detail-preserving One-step Diffusion Model for Multi-view Material Estimation

Recovering material information from images has been extensively studied in computer graphics and vision. Recent works in material estimation leverage diffusion model showing promising results. However, these diffusion-based methods adopt a multi-step denoising strategy, which is time-consuming for each estimation. Such stochastic inference also conflicts with the deterministic material estimation task, leading to a high variance estimated results. In this paper, we introduce StableIntrinsic, a one-step diffusion model for multi-view material estimation that can produce high-quality material parameters with low variance. To address the overly-smoothing problem in one-step diffusion, StableIntrinsic applies losses in pixel space, with each loss designed based on the properties of the material. Additionally, StableIntrinsic introduces a Detail Injection Network (DIN) to eliminate the detail loss caused by VAE encoding, while further enhancing the sharpness of material prediction results. The experimental results indicate that our method surpasses the current state-of-the-art techniques by achieving a $9.9\%$ improvement in the Peak Signal-to-Noise Ratio (PSNR) of albedo, and by reducing the Mean Square Error (MSE) for metallic and roughness by $44.4\%$ and $60.0\%$, respectively.

cs.CV↗

arXiv2025

Structural contribution to light-induced gap suppression in Ta$_2$NiSe$_5$

An excitonic insulator is a material that hosts an exotic ground state, where an energy gap opens due to spontaneous condensation of bound electron-hole pairs. Ta$_2$NiSe$_5$ is a promising candidate for this type of material, but the coexistence of a structural phase transition with the gap opening has led to a long-standing debate regarding the origin of the insulating gap. Here we employ MeV ultrafast electron diffraction to obtain quantitative insights into the atomic displacements in Ta$_2$NiSe$_5$ following photoexcitation, which has been overlooked in previous time-resolved spectroscopy studies. In conjunction with first-principles calculations using the measured atomic displacements, we find that the structural change can largely account for the photoinduced reduction in the energy gap without considering excitonic effects. Our work illustrates the importance of a quantitative reconstruction of individual atomic pathways during nonequilibrium phase transitions, paving the way for a mechanistic understanding of a diverse array of phase transitions in correlated materials where lattice dynamics can play a pivotal role.

cond-mat.mtrl-sci↗

arXiv2025

Super-resolution femtosecond electron diffraction reveals electronic and nuclear dynamics at conical intersections

Conical intersections play a pivotal role in excited-state quantum dynamics. Capturing transient molecular structures near conical intersections remains challenging due to the rapid timescales and subtle structural changes involved. We overcome this by combining the enhanced temporal resolution of mega-electron-volt ultrafast electron diffraction with a super-resolution real-space inversion algorithm, enabling visualization of nuclear and electronic motions at conical intersections with sub-angstrom resolution, surpassing the diffraction limit. We apply this technique to the textbook example of the ring-opening reaction of 1,3-cyclohexadiene, which proceeds through two conical intersections within 100 femtoseconds. The super-resolved transient structures near conical intersections reveal a C-C bond length difference of less than 0.4 angstrom and an approximately 30-femtosecond traversal time of the nuclear wave packet between them. These findings establish super-resolution ultrafast scattering as a transformative tool for uncovering quantum dynamics in molecules and open new avenues for studying light-matter interactions at the most fundamental level.

physics.chem-ph↗

arXiv2025

Motion-Aware Adaptive Pixel Pruning for Efficient Local Motion Deblurring

Local motion blur in digital images originates from the relative motion between dynamic objects and static imaging systems during exposure. Existing deblurring methods face significant challenges in addressing this problem due to their inefficient allocation of computational resources and inadequate handling of spatially varying blur patterns. To overcome these limitations, we first propose a trainable mask predictor that identifies blurred regions in the image. During training, we employ blur masks to exclude sharp regions. For inference optimization, we implement structural reparameterization by converting $3\times 3$ convolutions to computationally efficient $1\times 1$ convolutions, enabling pixel-level pruning of sharp areas to reduce computation. Second, we develop an intra-frame motion analyzer that translates relative pixel displacements into motion trajectories, establishing adaptive guidance for region-specific blur restoration. Our method is trained end-to-end using a combination of reconstruction loss, reblur loss, and mask loss guided by annotated blur masks. Extensive experiments demonstrate superior performance over state-of-the-art methods on both local and global blur datasets while reducing FLOPs by 49\% compared to SOTA models (e.g., LMD-ViT). The source code is available at https://github.com/shangwei5/M2AENet.

cs.CV↗

arXiv2025

Probing valence electron and hydrogen dynamics using charge-pair imaging with ultrafast electron diffraction

A key challenge in ultrafast science has been to directly track the coupled motions of electrons and nuclei in real-space and real-time. This study presents a significant step towards this goal by demonstrating the feasibility of time-resolved real-space tracking of valence electron and hydrogen dynamics during the photodissociation of ammonia (NH3) using MeV ultrafast electron diffraction. It is demonstrated that the enhanced temporal resolution, in conjunction with the analysis of the charge-pair distribution function, enables the disentanglement of the correlated motion of valence electrons and hydrogens in photoexcited ammonia molecule. The methodology employed in this study, which utilizes the charge-pair distribution function from ultrafast electron scattering to retrieve intertwined electron and nucleus dynamics, may open up new opportunities in the study of quantum dynamics for a wide range of molecules.

physics.chem-ph↗

arXiv2025

A Hybrid Prior Bayesian Method for Combining Domestic Real-World Data and Overseas Data in Global Drug Development

Background Hybrid clinical trial design integrates randomized controlled trials (RCTs) with real-world data (RWD) to enhance efficiency through dynamic incorporation of external data. Existing methods like the Meta-Analytic Predictive Prior (MAP) inadequately control data heterogeneity, adjust baseline discrepancies, or optimize dynamic borrowing proportions, introducing bias and limiting applications in bridging trials and multi-regional clinical trials (MRCTs). Objective This study proposes a novel hybrid Bayesian framework (EQPS-rMAP) to address heterogeneity and bias in multi-source data integration, validated through simulations and retrospective case analyses of risankizumab's efficacy in moderate-to-severe plaque psoriasis. Design and Methods EQPS-rMAP eliminates baseline covariate discrepancies via propensity score stratification, constructs stratum-specific MAP priors to dynamically adjust external data weights, and introduces equivalence probability weights to quantify data conflict risks. Performance was evaluated across six simulated scenarios (heterogeneity differences, baseline shifts) and real-world case analyses, comparing it with traditional methods (MAP, PSMAP, EBMAP) on estimation bias, type I error control, and sample size requirements. Results Simulations show EQPS-rMAP maintains estimation robustness under significant heterogeneity while reducing sample size demands and enhancing trial efficiency. Case analyses confirm superior external bias control and accuracy compared to conventional approaches. Conclusion and Significance EQPS-rMAP provides empirical evidence for hybrid clinical designs. By resolving baseline-heterogeneity conflicts through adaptive mechanisms, it enables reliable integration of external and real-world data in bridging trials, MRCTs, and post-marketing studies, broadening applicability without compromising rigor.

stat.ME↗

arXiv2025

Bi-directional Self-Registration for Misaligned Infrared-Visible Image Fusion

Acquiring accurately aligned multi-modal image pairs is fundamental for achieving high-quality multi-modal image fusion. To address the lack of ground truth in current multi-modal image registration and fusion methods, we propose a novel self-supervised \textbf{B}i-directional \textbf{S}elf-\textbf{R}egistration framework (\textbf{B-SR}). Specifically, B-SR utilizes a proxy data generator (PDG) and an inverse proxy data generator (IPDG) to achieve self-supervised global-local registration. Visible-infrared image pairs with spatially misaligned differences are aligned to obtain global differences through the registration module. The same image pairs are processed by PDG, such as cropping, flipping, stitching, etc., and then aligned to obtain local differences. IPDG converts the obtained local differences into pseudo-global differences, which are used to perform global-local difference consistency with the global differences. Furthermore, aiming at eliminating the effect of modal gaps on the registration module, we design a neighborhood dynamic alignment loss to achieve cross-modal image edge alignment. Extensive experiments on misaligned multi-modal images demonstrate the effectiveness of the proposed method in multi-modal image alignment and fusion against the competing methods. Our code will be publicly available.

cs.CV↗

arXiv2025

High-Frequency Prior-Driven Adaptive Masking for Accelerating Image Super-Resolution

The primary challenge in accelerating image super-resolution lies in reducing computation while maintaining performance and adaptability. Motivated by the observation that high-frequency regions (e.g., edges and textures) are most critical for reconstruction, we propose a training-free adaptive masking module for acceleration that dynamically focuses computation on these challenging areas. Specifically, our method first extracts high-frequency components via Gaussian blur subtraction and adaptively generates binary masks using K-means clustering to identify regions requiring intensive processing. Our method can be easily integrated with both CNNs and Transformers. For CNN-based architectures, we replace standard $3 \times 3$ convolutions with an unfold operation followed by $1 \times 1$ convolutions, enabling pixel-wise sparse computation guided by the mask. For Transformer-based models, we partition the mask into non-overlapping windows and selectively process tokens based on their average values. During inference, unnecessary pixels or windows are pruned, significantly reducing computation. Moreover, our method supports dilation-based mask adjustment to control the processing scope without retraining, and is robust to unseen degradations (e.g., noise, compression). Extensive experiments on benchmarks demonstrate that our method reduces FLOPs by 24--43% for state-of-the-art models (e.g., CARN, SwinIR) while achieving comparable or better quantitative metrics. The source code is available at https://github.com/shangwei5/AMSR

cs.CV↗

arXiv2025

DC-SAM: In-Context Segment Anything in Images and Videos via Dual Consistency

Given a single labeled example, in-context segmentation aims to segment corresponding objects. This setting, known as one-shot segmentation in few-shot learning, explores the segmentation model's generalization ability and has been applied to various vision tasks, including scene understanding and image/video editing. While recent Segment Anything Models have achieved state-of-the-art results in interactive segmentation, these approaches are not directly applicable to in-context segmentation. In this work, we propose the Dual Consistency SAM (DC-SAM) method based on prompt-tuning to adapt SAM and SAM2 for in-context segmentation of both images and videos. Our key insights are to enhance the features of the SAM's prompt encoder in segmentation by providing high-quality visual prompts. When generating a mask prior, we fuse the SAM features to better align the prompt encoder. Then, we design a cycle-consistent cross-attention on fused features and initial visual prompts. Next, a dual-branch design is provided by using the discriminative positive and negative prompts in the prompt encoder. Furthermore, we design a simple mask-tube training strategy to adopt our proposed dual consistency method into the mask tube. Although the proposed DC-SAM is primarily designed for images, it can be seamlessly extended to the video domain with the support of SAM2. Given the absence of in-context segmentation in the video domain, we manually curate and construct the first benchmark from existing video segmentation datasets, named In-Context Video Object Segmentation (IC-VOS), to better assess the in-context capability of the model. Extensive experiments demonstrate that our method achieves 55.5 (+1.4) mIoU on COCO-20i, 73.0 (+1.1) mIoU on PASCAL-5i, and a J&F score of 71.52 on the proposed IC-VOS benchmark. Our source code and benchmark are available at https://github.com/zaplm/DC-SAM.

cs.CV↗

arXiv2025

CKD: Contrastive Knowledge Distillation from A Sample-wise Perspective

In this paper, we propose a simple yet effective contrastive knowledge distillation framework that achieves sample-wise logit alignment while preserving semantic consistency. Conventional knowledge distillation approaches exhibit over-reliance on feature similarity per sample, which risks overfitting, and contrastive approaches focus on inter-class discrimination at the expense of intra-sample semantic relationships. Our approach transfers "dark knowledge" through teacher-student contrastive alignment at the sample level. Specifically, our method first enforces intra-sample alignment by directly minimizing teacher-student logit discrepancies within individual samples. Then, we utilize inter-sample contrasts to preserve semantic dissimilarities across samples. By redefining positive pairs as aligned teacher-student logits from identical samples and negative pairs as cross-sample logit combinations, we reformulate these dual constraints into an InfoNCE loss framework, reducing computational complexity lower than sample squares while eliminating dependencies on temperature parameters and large batch sizes. We conduct comprehensive experiments across three benchmark datasets, including the CIFAR-100, ImageNet-1K, and MS COCO datasets, and experimental results clearly confirm the effectiveness of the proposed method on image classification, object detection, and instance segmentation tasks.

cs.CV↗

arXiv2025

VTD-CLIP: Video-to-Text Discretization via Prompting CLIP

Vision-language models bridge visual and linguistic understanding and have proven to be powerful for video recognition tasks. Existing approaches primarily rely on parameter-efficient fine-tuning of image-text pre-trained models, yet they often suffer from limited interpretability and poor generalization due to inadequate temporal modeling. To address these, we propose a simple yet effective video-to-text discretization framework. Our method repurposes the frozen text encoder to construct a visual codebook from video class labels due to the many-to-one contrastive alignment between visual and textual embeddings in multimodal pretraining. This codebook effectively transforms temporal visual data into textual tokens via feature lookups and offers interpretable video representations through explicit video modeling. Then, to enhance robustness against irrelevant or noisy frames, we introduce a confidence-aware fusion module that dynamically weights keyframes by assessing their semantic relevance via the codebook. Furthermore, our method incorporates learnable text prompts to conduct adaptive codebook updates. Extensive experiments on HMDB-51, UCF-101, SSv2, and Kinetics-400 have validated the superiority of our approach, achieving more competitive improvements over state-of-the-art methods. The code will be publicly available at https://github.com/isxinxin/VTD-CLIP.

cs.CV↗

arXiv2025

BackMix: Regularizing Open Set Recognition by Removing Underlying Fore-Background Priors

Open set recognition (OSR) requires models to classify known samples while detecting unknown samples for real-world applications. Existing studies show impressive progress using unknown samples from auxiliary datasets to regularize OSR models, but they have proved to be sensitive to selecting such known outliers. In this paper, we discuss the aforementioned problem from a new perspective: Can we regularize OSR models without elaborately selecting auxiliary known outliers? We first empirically and theoretically explore the role of foregrounds and backgrounds in open set recognition and disclose that: 1) backgrounds that correlate with foregrounds would mislead the model and cause failures when encounters 'partially' known images; 2) Backgrounds unrelated to foregrounds can serve as auxiliary known outliers and provide regularization via global average pooling. Based on the above insights, we propose a new method, Background Mix (BackMix), that mixes the foreground of an image with different backgrounds to remove the underlying fore-background priors. Specifically, BackMix first estimates the foreground with class activation maps (CAMs), then randomly replaces image patches with backgrounds from other images to obtain mixed images for training. With backgrounds de-correlated from foregrounds, the open set recognition performance is significantly improved. The proposed method is quite simple to implement, requires no extra operation for inferences, and can be seamlessly integrated into almost all of the existing frameworks. The code is released on https://github.com/Vanixxz/BackMix.

cs.CV↗

arXiv2025

Efficient Masked AutoEncoder for Video Object Counting and A Large-Scale Benchmark

The dynamic imbalance of the fore-background is a major challenge in video object counting, which is usually caused by the sparsity of target objects. This remains understudied in existing works and often leads to severe under-/over-prediction errors. To tackle this issue in video object counting, we propose a density-embedded Efficient Masked Autoencoder Counting (E-MAC) framework in this paper. To empower the model's representation ability on density regression, we develop a new $\mathtt{D}$ensity-$\mathtt{E}$mbedded $\mathtt{M}$asked m$\mathtt{O}$deling ($\mathtt{DEMO}$) method, which first takes the density map as an auxiliary modality to perform multimodal self-representation learning for image and density map. Although $\mathtt{DEMO}$ contributes to effective cross-modal regression guidance, it also brings in redundant background information, making it difficult to focus on the foreground regions. To handle this dilemma, we propose an efficient spatial adaptive masking derived from density maps to boost efficiency. Meanwhile, we employ an optical flow-based temporal collaborative fusion strategy to effectively capture the dynamic variations across frames, aligning features to derive multi-frame density residuals. The counting accuracy of the current frame is boosted by harnessing the information from adjacent frames. In addition, considering that most existing datasets are limited to human-centric scenarios, we first propose a large video bird counting dataset, DroneBird, in natural scenarios for migratory bird protection. Extensive experiments on three crowd datasets and our \textit{DroneBird} validate our superiority against the counterparts. The code and dataset are available.

cs.CV↗

arXiv2025

Imaging the photochemical dynamics of cyclobutanone with MeV ultrafast electron diffraction

We study the photoinduced chemical dynamics of cyclobutanone upon excitation at 200 nm to the 3s Rydberg state using MeV ultrafast electron diffraction (UED). We observe both the elastic scattering signal, which contains information about the structural dynamics, and the inelastic scattering signal, which encodes information about the electronic state. Our results suggest a sub-picosecond timescale for the photodissociation dynamics, and an excited state lifetime of about 230 femtoseconds. The dissociation is found to be dominated by the C3 channel where cyclopropane and CO are produced. The branching ratio of the C3 channel to the C2 channel where ethene and ketene are produced, is estimated to be approximately 5:3. Our data suggest that the C3 and C2 channels account for approximately 80% of the photoproducts, with the remaining 20% exhibiting ring-opened structures. It is found that the timescale associated with the dissociation process in the C2 channel is shorter compared to that in the C3 channel. Leveraging the enhanced temporal resolution of MeV UED, our results provide a real-time mapping of the nuclear wavepacket dynamics, capturing the complete photochemical dynamics from S2 minimum through the S1/S0 conical intersection, and finally to the dissociation. Our experimental results provide new insights into the Norrish Type I reaction and can be used to benchmark non-adiabatic dynamics simulations.

physics.chem-ph↗

arXiv2025

Towards Robust Unsupervised Attention Prediction in Autonomous Driving

Robustly predicting attention regions of interest for self-driving systems is crucial for driving safety but presents significant challenges due to the labor-intensive nature of obtaining large-scale attention labels and the domain gap between self-driving scenarios and natural scenes. These challenges are further exacerbated by complex traffic environments, including camera corruption under adverse weather, noise interferences, and central bias from long-tail distributions. To address these issues, we propose a robust unsupervised attention prediction method. An Uncertainty Mining Branch refines predictions by analyzing commonalities and differences across multiple pre-trained models on natural scenes, while a Knowledge Embedding Block bridges the domain gap by incorporating driving knowledge to adaptively enhance pseudo-labels. Additionally, we introduce RoboMixup, a novel data augmentation method that improves robustness against corruption through soft attention and dynamic augmentation, and mitigates central bias by integrating random cropping into Mixup as a regularizer. To systematically evaluate robustness in self-driving attention prediction, we introduce the DriverAttention-C benchmark, comprising over 100k frames across three subsets: BDD-A-C, DR(eye)VE-C, and DADA-2000-C. Our method achieves performance equivalent to or surpassing fully supervised state-of-the-art approaches on three public datasets and the proposed robustness benchmark, reducing relative corruption degradation by 58.8% and 52.8%, and improving central bias robustness by 12.4% and 11.4% in KLD and CC metrics, respectively. Code and data are available at https://github.com/zaplm/DriverAttention.

cs.CV↗

arXiv2025

Dynamic Sub-graph Distillation for Robust Semi-supervised Continual Learning

Continual learning (CL) has shown promising results and comparable performance to learning at once in a fully supervised manner. However, CL strategies typically require a large number of labeled samples, making their real-life deployment challenging. In this work, we focus on semi-supervised continual learning (SSCL), where the model progressively learns from partially labeled data with unknown categories. We provide a comprehensive analysis of SSCL and demonstrate that unreliable distributions of unlabeled data lead to unstable training and refinement of the progressing stages. This problem severely impacts the performance of SSCL. To address the limitations, we propose a novel approach called Dynamic Sub-Graph Distillation (DSGD) for semi-supervised continual learning, which leverages both semantic and structural information to achieve more stable knowledge distillation on unlabeled data and exhibit robustness against distribution bias. Firstly, we formalize a general model of structural distillation and design a dynamic graph construction for the continual learning progress. Next, we define a structure distillation vector and design a dynamic sub-graph distillation algorithm, which enables end-to-end training and adaptability to scale up tasks. The entire proposed method is adaptable to various CL methods and supervision settings. Finally, experiments conducted on three datasets CIFAR10, CIFAR100, and ImageNet-100, with varying supervision ratios, demonstrate the effectiveness of our proposed approach in mitigating the catastrophic forgetting problem in semi-supervised continual learning scenarios.

cs.LG↗