- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Video Analysis and Summarization
- Human Pose and Action Recognition
- Virtual Reality Applications and Impacts
- Computer Graphics and Visualization Techniques
- Autonomous Vehicle Technology and Safety
- Advanced Image and Video Retrieval Techniques
- Generative Adversarial Networks and Image Synthesis
- Educational Technology and Pedagogy
- Robotic Path Planning Algorithms
- Data Mining Algorithms and Applications
- Parallel Computing and Optimization Techniques
- Human Motion and Animation
- Rough Sets and Fuzzy Logic
- Image and Video Stabilization
- Software System Performance and Reliability
- Embedded Systems Design Techniques
- Advanced Image Processing Techniques
- Anomaly Detection Techniques and Applications
- International Business and FDI
- Advanced Computational Techniques and Applications
- Hydraulic Fracturing and Reservoir Analysis
- Optical measurement and interference techniques
- Video Surveillance and Tracking Methods
Beihang University
2012-2025
Heilongjiang Provincial Institute of Hydraulic Research
2025
Northwest University
2021-2024
Shandong University of Finance and Economics
2024
Qingdao University of Technology
2022-2023
Tsinghua University
2013-2023
Harbin Institute of Technology
2023
State Key Laboratory of Virtual Reality Technology and Systems
2023
Ji Hua Laboratory
2023
Shandong University
2023
Video stabilization techniques are essential for most hand-held captured videos due to high-frequency shakes. Several 2D-, 2.5D-, and 3D-based have been presented previously, but the best of our knowledge, no solutions based on deep neural networks had proposed date. The main reason this omission is shortage in training data as well challenge modeling problem using networks. In paper, we present a video technique convolutional network. Previous works usually propose an off-line algorithm...
Example-guided image synthesis aims to synthesize an from a semantic label map and exemplary indicating style. We use the term "style" in this problem refer implicit characteristics of images, for example: portraits includes gender, racial identity, age, hairstyle; full body pictures it clothing; street scenes refers weather time day such like. A these cases indicates facial expression, pose, or scene segmentation. propose solution example-guided using conditional generative adversarial...
Inspired by the amino acid 2-chloro-4,5-dihydroxyphenylalanine (Cl-DOPA), present in composition of proteinaceous glue sandcastle worm Phragmatopoma californica, a simple strategy is presented to confer antifouling properties polymer surfaces using (but not releasing) bioinspired biocide. Cl-Dopamine used functionalize materials and hydrogel films easily, prevent biofilm formation on them.
This article examines the impact of inward Foreign Direct Investment (FDI) on host countries' domestic investment. Utilizing data from 50 countries over period 1970 to 2004, we find that FDI has a negative contemporaneous effect investment, while cumulative time tends be positive. In addition, separately study in Developed Countries (DCs) and Less (LDCs). The investment is DCs, neutral. Strong evidence suggests neutral LDCs,
Filling a small hole in an image with plausible content is well studied. Extrapolating to give distinctly larger one much more challenging---a significant amount of additional needed which matches the original image, especially near its boundaries. We propose data-driven approach this problem. Given source and direction(s) it be extrapolated, our system determines visually consistent for extrapolated regions using library images. As as considering low-level matching, we achieve consistency...
Virtual reality (VR) offers an artificial, computer generated simulation of a real life environment. It originated in the 1960s and has evolved to provide increasing immersion, interactivity, imagination, intelligence. Because deep learning systems are able represent compose information at various levels hierarchical fashion, they can build very powerful models which leverage large quantities visual media data. Intelligence VR methods applications been significantly boosted by recent...
Video portraits are common in a variety of applications, such as videoconferencing, news broadcasting, and virtual education training. We present novel method to synthesize photorealistic video for an input portrait video, automatically driven by person's voice. The main challenge this task is the hallucination plausible, facial expressions from speech audio. To address challenge, we employ parametric 3D face model represented geometry, expression, illumination, etc., learn mapping audio...
We introduce ShadowGAN, a generative adversarial network (GAN) for synthesizing shadows virtual objects inserted in images. Given target image containing several existing with shadows, and an input source object specified insertion position, the generates realistic shadow object. The is synthesized by generator; using proposed local global discriminators, synthetic shadow's appearance locally shape, globally consistent other objects' terms of direction area. To overcome lack training data,...
We present Write-A-Video , a tool for the creation of video montage using mostly text-editing. Given an input themed text and related repository either from online websites or personal albums, allows novice users to generate much more easily than current editing tools. The resulting illustrates given narrative, provides diverse visual content, follows cinematographic guidelines. process involves three simple steps: (1) user input, in form text, (2) automatically searches semantically...
With the recent rise of Metaverse, online multiplayer VR applications are becoming increasingly prevalent worldwide. However, as multiple users located in different physical environments, reset frequencies and timings can lead to serious fairness issues for collaborative/competitive applications. For apps/games, an ideal RDW strategy must make locomotion opportunities equal, regardless environment layouts. The existing methods lack scheme coordinate PEs, thus have issue triggering too many...
RGB-Infrared multi-modal object detection utilizes diverse and complementary information, showing some advantages in intelligent transportation field. The main challenge of is how to fuse the two modalities. difficulty fusion reflected aspects: 1) large visual differences between modalities make it difficult learn effective features, 2) misaligned images increase fusion. To this end, based on feature pyramid commonly used detection, we propose Multi-modal Feature Pyramid Transformer (MFPT)...
In numerous real-world applications, it is quite common that sample information partially available for some views due to machine breakdown or sensor failure, causing the problem of incomplete multi-view clustering (IMVC). While several IMVC approaches using view-shared anchors have successfully achieved pleasing performance improvement, (1) they generally construct with only one dimension, which could deteriorate diversity, bringing about serious loss; (2) constructed are typically a single...
Abstract Video stabilization is necessary for many hand‐held shot videos. In the past decades, although various video methods were proposed based on smoothing of 2D, 2.5D or 3D camera paths, hardly have there been any deep learning to solve this problem. Instead explicitly estimating and path, we present a novel online framework learn transformation each unsteady frame, given historical steady frames. Our network composed generative with spatial transformer networks embedded in different...
Scene text image super-resolution (STISR) has been regarded as an important pre-processing task for recognition from low-resolution scene images. Most recent approaches use the recognizer's feedback clues to guide super-resolution. However, directly using clue two problems: 1) Compatibility. It is in form of probability distribution, obvious modal gap with STISR - a pixel-level task; 2) Inaccuracy. it usually contains wrong information, thus will mislead main and degrade performance. In this...
Cattle behaviour is a significant indicator of cattle welfare. With the advancements in electronic equipment, monitoring and classifying multiple patterns becoming increasingly important precision livestock management. The aim this study was to detect physiological states using neural network model wearable sensors. A novel long short-term memory (LSTM) recurrent that uses two-way information developed accurately classify compared with baseline LSTM. Deep residual bidirectional LSTM were...
Vehicle-to-everything (V2X) autonomous driving opens up a promising direction for developing new generation of intelligent transportation systems. Collaborative perception (CP) as an essential component to achieve V2X can overcome the inherent limitations individual perception, including occlusion and long-range perception. In this survey, we provide comprehensive review CP methods scenarios, bringing profound in-depth understanding community. Specifically, first introduce architecture...
This work investigated the influence of cooling rate (0.04, 0.87, 15.2 and 156.2 K/s) on microstructure compressive properties Al-4Cu-3Li-0.7Mg-1Zn alloys fabricated by several casting molds. The experiment results revealed that cooled at low rates 0.04, 0.87 K/s is composed α-Al, α(Al)+T2, θ (Al2Cu), T1 phases core-shell configuration (Al13Fe4/Al7Cu2Fe phase). structure observed for first time in an Al-Cu-Li alloy. However, vanish high due to insufficient diffusion Cu, Li Fe elements. As...
Geographically dispersed users often rely on virtual avatars as intermediaries to facilitate interactive communication and collaboration. However, existing methods for augmented reality (AR) telepresence applications exhibit limitations, including restricted movement within confined sub-areas, lack of smooth transitions, the necessity manually establishing object mapping between dissimilar environments. We present a novel AR framework avatar locomotion adaption while preserving semantic...
This study experimentally investigates the effects of freezing conditions on shear characteristics geomembrane–soil interfaces, employing a temperature-controlled direct apparatus. The findings reveal significant variations in stress–shear displacement patterns at soil–geomembrane interface under different thermal conditions. At positive temperatures, manifests strain hardening behavior, whereas negative it transitions from weak softening low normal stress to strong high stress....
Abstract On 1 March 2018, President Trump declared a 25% tariff on certain steel imports by invoking Section 232 of the 1962 Trade Expansion Act. The pitted two America’s most storied and interconnected industries, auto producers, against one another made allies out longtime bitter political opponents Capitol Hill. Later that same year, doubled down when he initiated investigation parts imports. industry blasted proposal, while offered its strong support. This paper examines congressional...
The growing adoption of social virtual reality (VR) platforms underscores the importance safeguarding personal VR space to maintain user privacy and security. Teleportation, a prevalent instantaneous locomotion method in VR, facilitates engagement but can also inadvertently intrude upon space, thereby raising concerns. This paper introduces three innovative negotiated teleportation techniques designed secure user-to-user protect privacy, all under unified small-group development framework....