Lithography Hotspot Detection Model Based on Improved ShuffleNetV2 and Metric Learning
LI Zihao
XU Hui
Abstract：To address the insufficient feature extraction capability of lightweight models in lithography hotspot detection,a lithography hotspot detection model was proposed.This model which was named as ShuffleNetV2-MSDA-GHM-AAM(SMGA)adopted the improved shuffle net version 2(ShuffleNetV2)as its backbone network,incorporated a multi-scale dual attention(MSDA)module,and simultaneously integrated the gradient harmonizing mechanism(GHM)and additive angular margin(AAM)based on metric learning.The model's ability to model and perceive contextual information at different scales was enhanced,the inter-class discriminability of the feature embedding space was optimized,and dataset imbalance was alleviated.The experiments were conducted on the dataset of the 2012 international conference on computer-aided design(ICCAD 2012).The results showed that while maintaining a high detection recall of 98.22％,the average number of false alarms of the SMGA model was reduced to 484.This model provided a feasible solution for efficient and low-cost lithography hotspot detection in the integrated circuit design phase,possessing important engineering application value and promotion prospects.
Keywords：lithography hotspot detectionShuffleNetV2multi-scale attentionmetric learningadditive angular margingradient harmonizing mechanism
Publication Date：2025-12-30
Online Publishing Date：2025-12-10(First online date of this platform, not the publication date of the document)
Pages：6( 523-528 )