68‐3: A Unified Incremental Learning Object Detection Framework For Display Industry
SID Symposium Digest of Technical Papers · 2026-06 · 57 (1) · 941-944
DOI: 10.1002/sdtp.70001_257 ↗Kurz gefasst
Fachartikel in der Zeitschrift SID Symposium Digest of Technical Papers, online erschienen am 11.09.2026. Verfasst von Ting Wang und 6 weiteren Autor:innen. Die Arbeit nennt 11 Literaturangaben.
Abstract (Originalsprache)
Addressing the challenge of continuously emerging new defect categories in industrial defect detection where historical data cannot be retained, this paper proposes a unified incremental learning framework named MMA‐BiC‐AFD. The framework systematically addresses three major challenges—catastrophic forgetting, missing annotations, and class bias—through three core modules: Missing Annotation Modeling (MMA), Bias Correction (BiC), and Attention Feature Distillation (AFD). Evaluated on a real‐world OLED panel dataset, our method improves the average Precision from 82.5% to 91.5%, increases the average Recall from 81.0% to 90.5%, and significantly improves the old‐class Recall by 12%. While achieving state‐of‐the‐art performance, the proposed approach maintains a low computational cost of 86 GFLOPs, offering a lightweight and efficient solution for the continuous deployment of industrial AI systems.
Zitieren
APA
Wang, T., Wang, J., Chen, J., Li, Y., Kong, Y., Lv, B., & Jiang, X. (2026). 68‐3: A Unified Incremental Learning Object Detection Framework For Display Industry. SID Symposium Digest of Technical Papers, 57(1), 941-944. https://doi.org/10.1002/sdtp.70001_257
Vancouver
Wang T, Wang J, Chen J, Li Y, Kong Y, Lv B, et al. 68‐3: A Unified Incremental Learning Object Detection Framework For Display Industry. SID Symposium Digest of Technical Papers. 2026;57(1):941-944. doi:10.1002/sdtp.70001_257
BibTeX
@article{Wang2026,
author = {Wang, Ting and Wang, Jie and Chen, Jing and Li, Yunlong and Kong, Yue and Lv, Bin and Jiang, Xingqun},
title = {68‐3: A Unified Incremental Learning Object Detection Framework For Display Industry},
journal = {SID Symposium Digest of Technical Papers},
year = {2026},
volume = {57},
number = {1},
pages = {941--944},
doi = {10.1002/sdtp.70001_257}
}
EUVIMED