Chengfang Zhang

A lightweight knowledge reasoning method for large-scale knowledge graphs

Han ZHANG, Whenjun Zhou, B.L. Xu, Yuanbo Guo et autres

cn (code pays fourni par la source)

MDCFusion: Enhancing infrared and visible image fusion through Multi-Scale Dense Convolutional Sparse Coding

Junhao He, Chengfang Zhang, Yue Xiang Pan, Ziliang Feng

cn (code pays fourni par la source)

Improved Synthesis of Peach Fruit Moth Sex Pheromone and Evaluation of Field Trapping Effect

Yan Zheng, Jinhua Wang, Yubo Wang, Rongrong Fan et autres

The peach fruit moth is a significant pest of fruit crops. The primary component of its sex pheromone is (Z7)-eicosene-11-one (Z7-20-11kt). Using ethyl 4-bromobutyrate as the starting material, a Wittig reaction was conducted to synthesize ethyl (Z7)-undecenoate. This intermediate was subsequently hydrolyzed …

cn (code pays fourni par la source)

Enhanced visible-infrared pedestrian re-identification through feature decoupling with semi-orthogonal matrix

Shengyi Li, Chengfang Zhang, Xusong Ran

cn (code pays fourni par la source)

Preserving Identity in Portrait Video Generation via Facial Structure Injection

Zhenchun Liao, Dawei Dai, Le Yang, Chengfang Zhang

Although existing diffusion Transformer-based video generation models can synthesize high-quality dynamic portraits, they still face significant challenges in maintaining identity consistency across long sequences. We identify that the root cause lies in the model’s difficulty in decoupling stable identity features (such as …

cn (code pays fourni par la source)

Text-to-Face Generation Based on Human Preference

Le Yang, Zhenchun Liao, Qun Liu, Chengfang Zhang

Text-to-face image generation technology holds immense application value in fields such as virtual avatar creation, the entertainment industry, and public safety. However, while existing generative models can produce high-resolution images, they often overlook the subjective aesthetic information in text descriptions, leading to …

cn (code pays fourni par la source)

Multi-Granularity Masking Strategy for Self-Supervised Visual Representation Learning

Fenglin Yu, Shilin Zhao, Huan Gou, Shuyin Xia et autres

Masked autoencoders have recently emerged as a powerful paradigm for self-supervised visual representation learning by reconstructing missing image patches. However, the widely adopted random masking mechanism ignores the inherent structural organization of visual data, resulting in limited interpretability and poor integration of …

cn (code pays fourni par la source)

A Single-stage Interpretable Vision Transformer Model Via Granular-ball Computing

Huan Gou, Fenglin Yu, Guolai Jiang, Chengfang Zhang et autres

Deep learning models have been widely used in image recognition tasks due to their superior feature learning capabilities. The Vision Transformer (ViT), a mainstream deep learning architecture, uses a self-attention mechanism that effectively captures global feature dependencies compared to traditional convolutional neural …

cn (code pays fourni par la source)

Adapting Human Mesh Recovery with Vision-Language Feedback

Chongyang Xu, Buzhen Huang, Chengfang Zhang, Ziliang Feng et autres

Human mesh recovery can be approached using either regression-based or optimization-based methods. Regression models achieve high pose accuracy but struggle with model-to-image alignment due to the lack of explicit 2D-3D correspondences. In contrast, optimization-based methods align 3D models to 2D observations but …

Few-Shot Graph Anomaly Detection via Dual-Level Knowledge Distillation

Xuan Li, Chengfang Zhang, Ziliang Feng

Graph anomaly detection is crucial in many high-impact applications across diverse fields. In anomaly detection tasks, collecting plenty of annotated data tends to be costly and laborious. As a result, few-shot learning has been explored to address the issue by requiring only …

cn (code pays fourni par la source)

Digital Analysis of Natural Territorial Public Space Morphology Based on Spatial Syntax

Liwen Zhang, Yiyang Cao, Riyi Chen, Lujuan JIA et autres

Spatial syntax, as a quantifiable analytical tool, utilizes spatial scale division and segmentation to visually represent the morphology of natural territorial public spaces in a quantified graphic language, aiming to explore the relationship between public spaces and human behavioral activities. In order …

Unsupervised Restoration Method Based on Improved CycleGAN for Multiple Degenerate Images

Luobin Wu, Ziliang Feng, Zhaokang Guo, Chengfang Zhang

cn (code pays fourni par la source)