Shaojiang Yuan

Dr. Shaojiang Yuan

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 February 2025 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
Chenglizhao ChenChenglizhao Chen
Chenglizhao Chen
XLXinyu Liu
Xinyu Liu
MSMengke Song
Mengke Song
LLLuming Li
Luming Li
SYShaojiang Yuan
Shaojiang Yuan
SPShanchen Pang
Shanchen Pang
Video anomaly detection methods are mainly classified into two categories based on their primary feature types: appearance-based and action-based. Appearance-based methods rely on low-level visual features like color, texture, and shape, learning patterns specific to training scenes. While effective in familiar settings, they struggle with unknown or altered scenes due to poor generalization and limited understanding of action-scene relationships. In contrast, action-based methods focus on detecting action anomalies but often overlook contextual scene associations, leading to misjudgments (e.g., running on a street being deemed normal without considering scene context). To overcome these limitations, we propose a novel decoupling-based anomaly detection architecture (DecoAD). Its core lies in the decoupling and interweaving of scenes and actions, enabling explicit modeling of their complex relationships. By reconstructing these interactions using knowledge graphs, DecoAD achieves a deeper understanding of behaviors and contexts. This design ensures strong performance in both known and unknown scenarios, significantly enhancing generalization. To evaluate its effectiveness in dynamic scenes and its ability to handle scene-related anomalies, we introduce UFSR, the first video anomaly detection dataset featuring dynamic scenes and scene-related anomalies. DecoAD supports fully-supervised, weakly-supervised, and unsupervised settings, improving AUC on UBnormal by 1.1%, 3.1%, and 2.1% in fully-supervised, weakly-supervised, and unsupervised settings, and on UFSR by 1.2% and 8.2% in weakly-supervised and unsupervised settings. The source code and datasets are available at: https://github.com/liuxy3366/DecoAD.

30 August 2023 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
SYShaojiang Yuan
Shaojiang Yuan
LLLi Li
Li Li
HCHao Chen
Hao Chen
XLXian Li
Xian Li
Automated defect detection of reflective product surface is a challenging task due to the appearance of specular highlight. Although various highlight removal methods have been proposed, they fail to distinguish bright areas from highlight, and may remove bright defects as specular highlights, which can affect the accuracy of subsequent defect detection. In this paper, we propose a dual-mask guided deep learning model for detecting surface defects on highly reflective leather, which can remove surface specular highlights in industrial product images while preserving bright defects. Specifically, we propose an image transformation network for joint specular highlight detection and removal. The network is composed of multiple large window attention and MLP blocks, which can capture multi-scale features effectively. In addition, we create a dataset with 522 reflective leather images and perform pixel-level annotation of both high-light regions and defect regions. We conduct experiments on the reflective leather dataset and public dataset to evaluate the performance of our method. Compared with other approaches, the results demonstrate the superiority of our proposed method.

22 January 2024
Verified
Book Chapter
Closed access
PDF file
SYShaojiang Yuan
Shaojiang Yuan
LLLi Li
Li Li
NYNeng Yu
Neng Yu
TPTao Peng
Tao Peng
XHXinrong Hu
Xinrong Hu
XPXiong Pan
Xiong Pan

12 May 2025 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
SYShaojiang Yuan
Shaojiang Yuan
MSMengke Song
Mengke Song
JSJia Song
Jia Song
XLXinyu Liu
Xinyu Liu
Chenglizhao ChenChenglizhao Chen
Chenglizhao Chen
Knowledge distillation (KD) has demonstrated remarkable performance in the industrial anomaly detection (IAD) field, but its immense potential is constrained by the trade-off between the generalization of normal samples and the robustness of abnormal samples. High generalization often weakens the robustness against anomalies, leading to overlooked defects, while enhancing robustness can cause normality forgetting, where normal samples are misclassified as anomalies. This dilemma stems from the existing methods commonly assuming style invariance across normal and abnormal areas, focusing solely on the abnormal content information, while style information is usually neglected and put on the shelf. In this article, our main novelty lies in leveraging anomaly style shift estimation (SSE) for enlarging abnormal robustness while maintaining high normal generalization. This is the first attempt to introduce the style information into the KD-based IAD methods. By employing the bidirectional fusion of style and content information, our approach achieves more comprehensive and practical anomaly detection and localization results. Thus, our method can provide a more balanced and robust solution for the future IAD task. Experimental results on the MVTec2D, MVTec3D, BTAD, and KolektorSDD2 datasets demonstrate that our method outperforms major state-of-the-art (SOTA) methods in both accuracy and speed, with image-level and pixel-level area under the “Receiver Operating Characteristic” (AUROC) scores of 99.6% and 98.6% on MVTec2D, respectively. Furthermore, the efficiency of our method is validated in a real-world plastic crate defect detection system on a logistics production line.

15 November 2026 in Electronics
Verified
Journal Article
Open access
PDF file
SYShaojiang Yuan
Shaojiang Yuan
XLXiaoxue Lu
Xiaoxue Lu
CCChenglizhao Chen
Chenglizhao Chen
JSJia Song
Jia Song
Concealed dense prediction (CDP), including camouflaged object detection, RGB-D camouflaged object detection, and open-vocabulary camouflaged object segmentation, requires reasoning about objects whose appearance closely resembles their surroundings. Progress is constrained by the scarcity of diverse camouflage images with consistent pixel-level and multimodal annotations. Although generative models offer a promising source of training data, synthesizing useful camouflage supervision requires preserving object identity and scene geometry while adapting foreground appearance to the environment. Existing foreground-preserving synthesis pipelines often depend on input masks, whereas generic text-to-image generation does not explicitly coordinate camouflage-related attributes, object relations, and dense outputs. We present GenCAMO, a reference-guided framework for environment-aware camouflage image–dense annotation generation without manually supplied foreground masks at generation time. To support this task, we construct GenCAMO-DB, which enriches 34,200 images with text descriptions, depth maps, object attributes, and scene graphs through a semi-automatic annotation and verification pipeline. GenCAMO decouples scene-graph context into layout and semantic representations. A depth–layout coherence-guided ControlNet combines relational layout cues with depth information, while attribute-aware mask attention associates visual and textual features with the relevant objects and attributes. The resulting representations support camouflage image synthesis together with depth and mask estimation and subsequent refinement. Experiments evaluate both generation quality and synthetic-to-real transfer across three dense prediction tasks. GenCAMO achieves overall FID and KID scores of 49.69 and 0.0149, respectively, in the evaluated generation benchmark, and its synthetic data improves downstream camouflage prediction. Ablation experiments and a user study further support the value of explicit contextual modeling for generating useful camouflage supervision.