A5101925114- 3D Shape Modeling and Analysis
- Computer Graphics and Visualization Techniques
- Advanced Vision and Imaging
- Human Pose and Action Recognition
- Industrial Vision Systems and Defect Detection
- Wood Treatment and Properties
- Advanced Image and Video Retrieval Techniques
- Advanced Sensor and Energy Harvesting Materials
- Scientific Computing and Data Management
- Advanced Neural Network Applications
- Human Motion and Animation
- ECG Monitoring and Analysis
- Image Processing and 3D Reconstruction
- Magnetic Properties and Synthesis of Ferrites
- Wood and Agarwood Research
- Cardiovascular Health and Disease Prevention
- Tactile and Sensory Interactions
- EEG and Brain-Computer Interfaces
- Electromagnetic wave absorption materials
- Generative Adversarial Networks and Image Synthesis
- Image Enhancement Techniques
- Advanced Graph Neural Networks
- Image and Object Detection Techniques
- Phonocardiography and Auscultation Techniques
- Advanced Materials and Mechanics
University of Electronic Science and Technology of China
2022-2024
Southwest Forestry University
2021-2024
Hangzhou Dianzi University
2024
Shenzhen University
2024
Hubei University of Technology
2022-2024
ShanghaiTech University
2021-2023
Anhui Jianzhu University
2022-2023
Jiangsu University
2016-2023
Jiangsu Normal University
2023
Institute of Automation
2022
Recent CLIP-guided 3D optimization methods, such as DreamFields [19] and PureCLIPNeRF [24], have achieved impressive results in zero-shot text-to-3D synthesis. However, due to scratch training random initialization without prior knowledge, these methods often fail generate accurate faithful structures that conform the input text. In this paper, we make first attempt introduce explicit shape priors into process. Specifically, a high-quality from text text-to-shape stage prior. We then use it...
Abstract The nondestructive testing technology of generated acoustic emission (AE) signals for wood is great significance the evaluation internal damages wood. To achieve more accurate and adaptive evaluation, an AE classification method combining empirical mode decomposition (EMD), discrete wavelet transform (DWT), linear discriminant analysis (LDA) classifier proposed. Five features (entropy, crest factor, pulse margin waveform factor) are selected because they sensitive to uncertainty,...
Transformers have shown preferable performance on many vision tasks. However, for the task of person re-identification (ReID), vanilla transformers leave rich contexts high-order feature relations under-exploited and deteriorate local details, which are insufficient due to dramatic variations pedestrians. In this work, we propose an Omni-Relational High-Order Transformer (OH-Former) model omni-relational features ReID. First, strengthen capacity visual representation, instead obtaining...
Existing view synthesis methods mainly focus on the perspective images and have shown promising results. How-ever, due to limited field-of-view of pinhole cam-era, performance quickly degrades when large cam-era movements are adopted. In this paper, we make first attempt generate novel views from a single indoor panorama take camera translations into consideration. To tackle challenging problem, use Convolutional Neural Networks (CNNs) extract deep features estimate depth map source-view...
We present a high-power, polarization-maintaining (PM) Er <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3+</sup> -doped fluoride fiber amplifier, which enables watt- level, linearly-polarized (LP) superfluorescent outputs around 2.8 μm by using non-PM source (FSS) after polarization controlling as the seed. Based on an optimized scheme, gained signal with 20 dB bandwidth of 95 nm (2710∼2805 nm) can yield maximum power 2.32 W and extinction...
We describe the value of reframing digital curation practices through a computational thinking (CT) framework. Using case study that demonstrates treatments World War II Japanese-American Incarceration Camp Records, we demonstrate applicability CT with respect to: (1) Detecting personally identifiable information, (2) Developing name registries, (3) Integrating vital records, (4) Designing controlled vocabularies, (5) Mapping events and people, (6) Connecting people networks. The work was...
Enhancing AI systems to perform tasks following human instructions can significantly boost productivity. In this paper, we present InstructP2P, an end-to-end framework for 3D shape editing on point clouds, guided by high-level textual instructions. InstructP2P extends the capabilities of existing methods synergizing strengths a text-conditioned cloud diffusion model, Point-E, and powerful language models, enabling color geometry using To train introduce new dataset, constructed integrating...
In this paper we presented a novel idea for large-scale 3D scene reconstruction and annealing based image grouping algorithm unordered wide-baseline photos. Firstly, an alternative maximum posterior (MAP) model which can easily incorporate clustering prior knowledge is proposed. Second, efficient developed organizing photos into clusters by calculating matching number of invariant features. Thirdly, analyze the time complexity efficiency proposed approach. Finally series experiments are...
Just like data plays a fundamental role in perceptual intelligence, simulated environments become the cornerstone of artificial agents cognitive intelligence. Meanwhile, evaluation decision has put forward an urgent need for human-machine accessible game platforms, on which humans and can interact confront, even "Turing test" be conducted. In this work, MiaoSuan wargame platform, human-computer gaming platform multi-agent imperfect information is proposed. It provides multiple scenes tasks,...
Abstract The nondestructive testing technology of generated acoustic emission(AE) signals for wood is great significance the evaluation internal damages wood. In order to improve classification accuracy and adaptability AE signal, we selected two features(pseudospectrum, entropy) classify in process fracture using SVM classifier. three-point bending load damage experiment was utilized generate original signals. Evaluation indexes(Precision, Accuracy, Recall, F1-score, Cohen Kappa score,...
This paper aims to study the mechanical properties of different graded concrete aggregates and discuss influence performance. Through literature review, research status aggregate gradation at home abroad is summarized, shortcomings existing studies are summarized. On this basis, a variety selected prepare specimens, their tested, including compressive strength, folding strength other indexes. The experimental results show that significantly different, deformation matched have important...
We present InstantMesh, a feed-forward framework for instant 3D mesh generation from single image, featuring state-of-the-art quality and significant training scalability. By synergizing the strengths of an off-the-shelf multiview diffusion model sparse-view reconstruction based on LRM architecture, InstantMesh is able to create diverse assets within 10 seconds. To enhance efficiency exploit more geometric supervisions, e.g, depths normals, we integrate differentiable iso-surface extraction...
We introduce NovelGS, a diffusion model for Gaussian Splatting (GS) given sparse-view images. Recent works leverage feed-forward networks to generate pixel-aligned Gaussians, which could be fast rendered. Unfortunately, the method was unable produce satisfactory results areas not covered by input images due formulation of these methods. In contrast, we novel view denoising through transformer-based network 3D Gaussians. Specifically, incorporating both conditional views and noisy target...
Recent advancements in generative models have significantly improved novel view synthesis (NVS) from multi-view data. However, existing methods depend on external alignment processes, such as explicit pose estimation or pre-reconstruction, which limits their flexibility and accessibility, especially when is unstable due to insufficient overlap occlusions between views. In this paper, we propose NVComposer, a approach that eliminates the need for alignment. NVComposer enables model implicitly...