Bingqian Li
Research Subjects
Bingqian Li has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Baoru Ge
School of Fashion, Henan Institute of Science and Technology, Xinxiang 453000, China
1 shared publication - Haojie Chai1 shared publication
- Shuhua Zhang1 shared publication
- Yangyang Zhu1 shared publication
Latest Publications
- Open Access
- 1 May 2026
- Published by Elsevier in SSRN Electronic Journal
Digital Extraction of Decorative Motifs on Women's Jackets from the Late Qing Dynasty and Early Republican Period Using a Dense Feature Aggregation Network
Decorative motifs on women's jackets from the late Qing dynasty and early Republican period hold significant cultural and artistic value. Their accurate digital extraction is therefore essential for preserving traditional garments. However, existing approaches, including manual extraction and conventional algorithms, are often hindered by low efficiency, imprecise boundary localization, and limited robustness in complex visual contexts. To address these limitations, a two-stage edge detection framework based on a dense feature aggregation network is proposed. Specifically, a bidirectional hierarchical fusion decoder effectively integrates global semantic information with fine-grained structural details, enabling the generation of high-resolution and pixel-accurate edge maps. In addition, a multi-scale attention-guided module is introduced to adaptively fuse multi-level edge predictions, which suppresses background interference and improves contour continuity and clarity. Experimental results on a self-constructed dataset demonstrate that the method achieves an Optimal Dataset Scale (ODS) of 0.865, an Optimal Image Scale (OIS) of 0.878, and an Average Precision (AP) of 0.895, outperforming several main edge detection approaches. The approach produces fine-grained, continuous, and structurally consistent motif representations, providing a practical approach for the digital preservation and dissemination of traditional decorative heritage.
- 1 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDigital Extraction of Decorative Motifs on Women's Jackets from the Late Qing Dynasty and Early Republican Period Using a Dense Feature Aggregation Network
Decorative motifs on women's jackets from the late Qing dynasty and early Republican period hold significant cultural and artistic value. Their accurate digital extraction is therefore essential for preserving traditional garments. However, existing approaches, including manual extraction and conventional algorithms, are often hindered by low efficiency, imprecise boundary localization, and limited robustness in complex visual contexts. To address these limitations, a two-stage edge detection framework based on a dense feature aggregation network is proposed. Specifically, a bidirectional hierarchical fusion decoder effectively integrates global semantic information with fine-grained structural details, enabling the generation of high-resolution and pixel-accurate edge maps. In addition, a multi-scale attention-guided module is introduced to adaptively fuse multi-level edge predictions, which suppresses background interference and improves contour continuity and clarity. Experimental results on a self-constructed dataset demonstrate that the method achieves an Optimal Dataset Scale (ODS) of 0.865, an Optimal Image Scale (OIS) of 0.878, and an Average Precision (AP) of 0.895, outperforming several main edge detection approaches. The approach produces fine-grained, continuous, and structurally consistent motif representations, providing a practical approach for the digital preservation and dissemination of traditional decorative heritage.