计算机科学
薄膜晶体管
过程(计算)
分割
推论
人工智能
晶体管
半导体器件制造
方案(数学)
计算机硬件
生产(经济)
资源(消歧)
炸薯条
图像分割
电子工程
制造工艺
液晶显示器
实时计算
计算机视觉
市场细分
可扩展性
吞吐量
生产线
图像处理
嵌入式系统
产品(数学)
光学接近校正
数据建模
贴片设备
作者
Alin Hou,Junpeng Zhang,Hai Gong,Jie Chen,Jiayu Sun,Minjun Xu,Mengxue Yan,Wen Cao
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2025-04-25
卷期号:: 51-55
标识
DOI:10.1109/ic-c68228.2025.00021
摘要
This paper presents a comprehensive study on the application and optimization of the DeepLabV3+ model for pixel-level semantic segmentation in ultra-high resolution Thin-Film Transistor (TFT) Liquid Crystal Display (LCD) images. The research addresses two critical challenges: the inability of conventional models to directly process ultra-high-resolution TFT images and the excessive memory and GPU resource consumption associated with such images. The goal is to demonstrate the feasibility and efficiency of the DeepLabV3+ model in accurately segmenting large-scale TFT images, thereby improving the automated optical inspection process in TFT manufacturing and ensuring product quality. The proposed model enables direct training and inference on ultra-high-resolution images, seamlessly integrating into the automated optical inspection (AOI) of TFT panels. This optimization significantly reduces inspection time, improves efficiency, and ultimately enhances production throughput.
科研通智能强力驱动
Strongly Powered by AbleSci AI