Jingdong Zhang

MTPano: Multi-Task Panoramic Scene Understanding via Label-Free Integration of Dense Prediction Priors

Jingdong Zhang, Xiaohang Zhan, Lingzhi Zhang, Yizhou Wang et autres

Comprehensive panoramic scene understanding is critical for immersive applications, yet it remains challenging due to the scarcity of high-resolution, multi-task annotations. While perspective foundation models have achieved success through data scaling, directly adapting them to the panoramic domain often fails due to …

us (code pays fourni par la source)

SPGen: Spherical Projection as Consistent and Flexible Representation for Single Image 3D Shape Generation

Jingdong Zhang, Weikai Chen, Jionghao Wang, Zhengming Yu et autres

Existing single-view 3D generative models typically adopt multiview diffusion priors to reconstruct object surfaces, yet they remain prone to inter-view inconsistencies and are unable to faithfully represent complex internal structure or nontrivial topologies. In particular, we encode geometry information by projecting it …

us, ru, hk (code pays fourni par la source)

High-Speed Target Tracking and Fusion Estimation Method Based on Adaptive Improved Cubature Kalman Filter

Jinbao Zhang, Haiguang Li, Jingdong Zhang, Boyang Liu et autres

Hypersonic target tracking is a major challenge currently faced by estimation theory, with its core difficulties lying in the strong nonlinear dynamic model, uncertain time-varying noise statistical characteristics, and complex maneuvering modes. Traditional nonlinear filtering algorithms often struggle to balance estimation accuracy, …

cn (code pays fourni par la source)

Multi-Task Label Discovery via Hierarchical Task Tokens for Partially Annotated Dense Predictions

Jingdong Zhang, Hanrong Ye, Xin Li, Wenping Wang et autres

In recent years, simultaneous learning of multiple dense prediction tasks with partially annotated label data has emerged as an important research area. Previous works primarily focus on leveraging cross-task relations or conducting adversarial training for extra regularization, which achieve promising performance improvements, …

us, hk (code pays fourni par la source)

BridgeNet: Comprehensive and Effective Feature Interactions via Bridge Feature for Multi-Task Dense Predictions

Jingdong Zhang, Jiayuan Fan, Peng Ye, Bo Zhang et autres

Multi-task dense prediction aims at handling multiple pixel-wise prediction tasks within a unified network simultaneously for visual scene understanding. However, cross-task feature interactions of current methods are still suffering from incomplete levels of representations, less discriminative semantics in feature participants, and inefficient …

cn, us (code pays fourni par la source)

A data backup optimization mechanism for the virtual FPGA cluster

Wei Liu, Fang Yu, Jingdong Zhang, Yanmei Shen et autres

FESSNC: Fast Exponentially Stable and Safe Neural Controller

Jingdong Zhang, Qunxi Zhu, Wei Lin

In order to stabilize nonlinear systems modeled by stochastic differential equations, we design a Fast Exponentially Stable and Safe Neural Controller (FESSNC) for fast learning controllers. Our framework is parameterized by neural networks, and realizing both rigorous exponential stability and safety guarantees. …

More Efficient and Locally Enhanced Transformer

Zhefeng Zhu, Ke Qi, Yicong Zhou, Wenbin Chen et autres

cn, mo (code pays fourni par la source)

Template Shift and Background Suppression for Visual Object Tracking

Yiwei Wu, Ke Qi, Wenbin Chen, Jingdong Zhang et autres

cn (code pays fourni par la source)

Rethinking Cross-Domain Pedestrian Detection: A Background-Focused Distribution Alignment Framework for Instance-Free One-Stage Detectors

Yancheng Cai, Bo Zhang, Baopu Li, Tao Chen et autres

Cross-domain pedestrian detection aims to generalize pedestrian detectors from one label-rich domain to another label-scarce domain, which is crucial for various real-world applications. Most recent works focus on domain alignment to train domain-adaptive detectors either at the instance level or image level. …

gb, cn, us (code pays fourni par la source)

Image Recognition Based on Enhanced-Conformer

Runlin Gong, Ke Qi, Yicong Zhou, Wenbin Chen et autres

Convolutional Neural Networks (CNNs) has always dominated visual recognition tasks, and it is difficult to link distant information in images due to the size limitation of each convolution filter. Vision Transformer (ViT) can capture features at a distance in an image, but …

cn, mo (code pays fourni par la source)

A low-overhead RO PUF design for Xilinx FPGAs

Songwei Pei, Jingdong Zhang, Ruonan Wang

Ring Oscillator (RO) Physical Unclonable Function (PUF) can effectively generate unique chip responses to support a variety of security-related applications. However, RO PUF typically incurs high hardware overhead when implemented in FPGA. In this paper, we designed a low-overhead RO PUF for …

cn (code pays fourni par la source)