Video virtual try-on (VTON) aims to dress a target person in a desired garment while preserving the motion and identity in the original video. Generating long-duration VTON videos exacerbates challenges in achieving temporal coherence and visual fidelity. Existing image-based methods struggle with temporal consistency due to their frame-by-frame manner, while video-based approaches often sacrifice details for global coherence, resulting in blurry results. To address these limitations, we propose the first VTON framework tailored for long-video generation, namely LV-VTON, based on UNet Diffusion Transformer (UDiT) and Visual Autoregressive Modeling (VAR). Our LV-VTON framework includes three key components, i.e., Garment Appearance Module, Temporal Consistency Module, and Extended Video Synthesis Module. Notably, we collected a diverse and high-quality dataset, namely LongTry, to advance long video virtual try-on research. Extensive experiments demonstrate that LV-VTON excels at synthesizing various long VTON videos, outperforming state-of-the-art methods in detail preservation, temporal consistency, and long-video synthesis.

Character animation aims to generate realistic, high-quality videos from a reference image and target frames. However, existing methods struggle to balance fine-grained detail preservation with temporal consistency. This limitation results in artifacts like flickering and unrealistic deformations, especially in facial and hand regions. To address these challenges, we propose DreamAnimate, a novel framework that synthesizes temporally consistent and detail-rich animations. DreamAnimate integrates three modules: the Progressive Motion Estimation module ensures accurate motion alignment and temporal stability by refining keypoint heatmaps, the Global Affine Transformation module generates dense motion flows to handle complex motions and occlusions, and the Character Animation Fusion module combines intermediate synthesis using a UNet architecture and an Animation Fusion Network to produce high-quality animations. Extensive experiments demonstrate that DreamAnimate outperforms state-of-the-art methods, achieving superior fidelity and effectively capturing intricate facial expressions and hand movements. Code and models will be released at https://github.com/cslltian/DreamAnimate in the near future.

Visual emotion recognition (VER), which aims to understand human emotional reactions to different visual stimuli, has garnered increasing attention. However, the inherent ambiguity of emotional features presents significant challenges for data annotation in supervised learning paradigms. To address this limitation, emotion domain adaptation (EDA) facilitates knowledge transfer from labeled source domains to unlabeled target domains. Recently, large visual-language models such as CLIP have demonstrated impressive transfer performance on traditional UDA tasks. However, when generalizing to more abstract concepts such as emotion, the misalignment between CLIP and emotion spaces greatly affects the model performance. To address these challenges, we propose a CLIP-based emotion disentanglement (EmoD) framework designed for EDA. Leveraging perspectives from information bottleneck theory, EmoD implements a disentangler network that extracts emotion-specific features while removing redundant emotion-agnostic information. It also incorporates cross-domain feature alignment to reduce the affective gap between domains. Experimental evaluations in six EDA settings demonstrate that EmoD achieves state-of-the-art performance, surpassing traditional CLIP-based UDA methods by an average of 2.53%.

Stainless steel/carbon steel double-layer structures are commonly used in industries, but they are prone to generate internal defects (such as delamination and corrosions) at the bonding interface of the carbon steel layer. Eddy current pulsed thermography (ECPT) has shown promise for subsurface defect detection due to its high excitation and concentrated heating area. However, the complex heat transfer in multilayer structures lead to poor signal-to-noise ratios and poses challenges for defect identification. Moreover, the data-driven algorithms like PCA and ICA, though widely used in postprocessing, are faced with unstable performance and poor interpretability due to the lack of attention to the specific physical mechanism. To address these issues, this article proposes a thermal parameter reconstruction (TPR) imaging method to better detect the defect regions. Specifically, TPR regards the test sample as a 3-D thermal impedance space and projects it onto a parameterized grid plane. According to the thermal diffusion mechanism in ECPT, TPR builds a physical model to calculate the spatial thermal parameters of the projected surface. Through the reconstructed visual thermal parameters grid, the shape information of internal defects can be more clearly detected. An experiment on double-layer stainless steel/carbon steel structures is conducted, which validates the enhancement effectiveness of TPR on both round and irregularly shaped interlayer defects.

Insulated gate bipolar transistors (IGBTs) are the key semiconductor power devices in the systems of power electronics due to the advantages of large capacity, fast switching speed, easy to drive, low on-voltage and high input impedance. Due to the temperature swinging and mismatches of the coefficients of thermal expansion (CTE) of internal materials in IGBT modules, electrical-thermal-mechanical coupling is generated and inevitably results in thermal deformation. This thermal deformation is one of the key factors causing IGBT reliability issues. Therefore, it is of great significance to study the electrical-thermal-mechanical deformation characteristics of IGBT modules.

Switching mechanical wave (SMW) effect has attracted extensive attention over the past few years owing to its broad application prospects in condition monitoring for insulated gate bipolar transistor (IGBT) devices. The latest research suggests that the method of directly measuring the mechanical vibration signal on IGBT bare die by laser vibrometer is of great significance and potential to the mechanism revelation of SMW effect and the accurate extraction of its physical characteristics. Nevertheless, the measured vibration signals in IGBT modules by laser vibrometer typically contain a variety of other interference signals, such as thermal deformation signal, electromagnetic jamming, and background noise, in addition to the expected target SMW signals. Therefore, the aim of this work is to develop an effective signal decoupling method to realize the extraction of SMW signal from the aliasing signals obtained by laser vibration measurement.

Switching mechanical wave (SMW) occurring at switching moment in insulated gate bipolar transistor (IGBT) has attracted extensive attention in recent years. However, AE sensors widely utilized in current researches can hardly realize the SMW detection in the local area of single-tube IGBTs due to the limitations of large volume and low spatial resolution, thus hindering the further investigation of SMW effect mechanism revelation and physical properties. Therefore, this work aims to develop a high spatial resolution laser interferometric detection system to capture SMW signals at different spatial locations of single-tube IGBTs, thus enhancing the comprehension on the underlying mechanisms and physical properties of SMW effect in single-tube IGBTs.

Eddy current pulsed thermography (ECPT) has attracted much attention in nondestructive testing for its noncontact and large field of view. However, the ECPT images usually suffer from the thermal diffusion blurs. The fusion of temperature spatial and temporal features is the hotspot in the new research of ECPT enhancing methods, but these two features are contradictory on the speed and the accuracy of algorithms. Specifically, spatial features process fast but perform poorly in accuracy and antinoise ability, while temporal features are usually calculated from the whole ECPT video sequence, which inevitably increases the demand for data storage and the time cost, especially in the detection of large workpieces, such as engine blades or pressure pipelines. In this article, an enhanced diffusion-based method (EDBM) is proposed to solve this issue, which maps the temporal features through the spatial features of a single ECPT image, significantly reduces the input data volume and shows great potential in ECPT online detection. Experiments on multiple artificial and natural samples verify that, compared with raw ECPT image, the proposed EDBM can reduce root-mean-square error by 51.7%–86.5% (73.2% on average) and improve signal-to-noise ratio by 3.70–16.8 times (6.82 times on average), which performs better than the commonly spatial-based and temporal-based ECPT enhancement algorithms, such as enhanced Canny and independent component analysis, close to the latest sparse-model decomposition methods, but with two orders of magnitude less time cost.

The recently proposed induced current thermo-electrical impedance tomography (ICTEIT) is a nondestructive evaluation method for nonferromagnetic metal materials. The method reconstructs the defect profile by solving conductivity distribution from eddy currents, which has shown good performance in simple defects, but worse performance in complex defects (such as multiple adjacent defects and defects with complex contours). The main reason is that the unidirectional excitation produces low eddy current regions near the defect, which lacks reliable reconstruction information. For the problem, multidirectional excitation is able to make up for the low eddy current regions. However, previous methods simply superimpose the information of all eddy currents, and cannot integrate them for reconstruction. To solve the problem, we present a multidirectional information fusion method for defect reconstruction. The proposed method uses least squares optimization and transforms the information of multidirectional currents into a system of linear equations associated with conductivity, which constrains the solution and makes the solution satisfy each current simultaneously. Furthermore, to solve the ill-conditioning in the inverse problem, the total variation regularization method is introduced. Experiments are conducted on multiple neighboring defects and defects with complex shapes to validate the proposed method's performance.

Nonferromagnetic metal materials are widely used in industry. Defects generated during manufacture and use may lead to serious accidents. The defect reconstruction is important for nondestructive evaluation. Eddy current pulsed thermography (ECPT) is a well-known nondestructive testing and evaluation method, but hardly reconstruct fully defect profile. Electrical impedance tomography (EIT) shows promising potential in defect profile reconstruction, but suffers from electrode number limitation. In this article, EIT is introduced to ECPT image sequences processing, and a new method, induced current thermo-electrical impedance tomography, is developed. The proposed method takes each infrared camera pixel as a virtual electrode, captures the electric current distribution with spatial resolution as high as camera, and reconstructs conductivity distribution at pixel level from which the defect profile can be identified. Experiments with different defects are carried out to verify the performance of the proposed method.

Deepfake technique can synthesize realistic images, audios, and videos, facilitating the thriving of entertainment, education, healthcare, and other industries. However, its abuse may pose potential threats to personal privacy, social stability, and even national security. Therefore, the development of deepfake detection methods is attracting more and more attention. Existing works mainly focus on the detection of common videos for entertainment purposes. In contrast, fake videos maliciously synthesized for Person of Interest (PoI, i.e., who is in an authoritative position and has broadly public influences) are much more harmful to society because of celebrity endorsement. However, there is no particular benchmark for driving related research in the community. Motivated by this observation, we present the first large-scale benchmark dataset, named FakePoI, to enable the research on fake PoI detection. It contains numerous fake videos of important people from all walks of life, e.g., police chiefs, city mayors, famous artists, and well-known Internet bloggers. In summary, our FakePoI includes 11092 synthesized videos where only a few clips rather than the entire are fake. Previous fake detection algorithms deteriorate heavily or even fail on our FakePoI due to two main challenges. On the one hand, the rich diversity of our fake videos makes it pretty difficult to find universally applicable patterns for detection. On the other hand, the high credibility contributed by the presence of real frames easily confuses a common detector. To tackle these challenges, we present an amplifier framework, highlighting the feature gap between real and generated video frames. Specifically, we present a quadruplet loss to narrow the distance of all real PoIs and meanwhile push away each real and fake PoI in embedding space. We implement our framework and conduct extensive experiments on the proposed benchmark. The quantitative results demonstrate that our approach outperforms existing methods significantly, setting a strong baseline on FakePoI. The qualitative analysis also shows its superiority. We will release our dataset and code athttps://github.com/cslltian/deepfake-detectionto encourage future research on this valuable area.

This paper presents an investigation of dynamic thermal deformation characteristics in insulated gate bipolar transistor (IGBT) modules under bonding wire cracking condition by means of finite element simulation and experimental validation. Firstly, a realistically restored three-dimensional geometric model for IGBT modules is constructed and simulated to investigate thermal deformation field. Then the thermal deformation field characteristics under bonding wire intact and cracked conditions are compared and analyzed indepth. The result shows that the thermal deformation fluctuation amplitude of the cracked bonding wire decreases by 82%, while the thermal deformation value of other unbroken wires increases by 35% on average. Finally, the experimental verification is carried out, and the conclusion shows that it coincides well with the simulation results. This work provides confident evidence and important data to facilitate more precise life-time predictions and thermal-mechanical reliability assessment for power electronic modules.