Friday, May 2025

09:20 AM - 09:40 AM

Room: LL21CD

Session: Artificial Intelligence / Machine Learning

Defect Classification Algorithms for Display Manufacturing Based on the Convolutional Neural Network Mixture-of-Experts Model

• Yunlong Li

  BOE Technology Group Co., Ltd.

Description:

This paper proposed a sparse network structure to classify defects based on the method of mixing pre-expert models: the model fuses the feature extracted by the pre-expert model through the gating unit, and assigns the fused features to different feature encoding units through the gated feature separation unit. The mixture-of-experts layer is used to extract and transform the semantic features, and finally the prediction results are obtained through the sparse decoding structure. In order to prevent over-fitting caused by a small amount of data, this paper improves the convergence performance of the model and reduces the computational cost by reusing part of the structure of the pre-expert to improve the generalization of feature extraction.