Session and scheduling information are listed below. Select a session from the list and press "Go" to view the abstracts for that session.
Session
32
: AI/ML for Display Manufacturing |
Display Manufacturing
; Artificial Intelligence Including Machine Learning for Imaging
|
Wednesday, May 15 / 09:00 AM - 10:00 AM / San Jose Convention Center, LL20A
Chair:
Hyoungsik Nam, Kyung Hee University, Seoul, South Korea
Co-Chair:
Daniel Lee, AU Optronics Corp, Hsinchu, Taiwan Roc
32.1 - Improving QD Backplane Defect Image Generation Using Automatic Masking in Diffusion Models (9:00 AM - 9:20 AM)
-
Zhihong Pan, Rahul Shenoy, Kaushik Balakrishnan, Qisen Cheng, Janghwan Lee
Samsung Display America Lab San Jose CA US
-
Yongmoon Jeon, Deokyeong Jeong, Jaewon Kim
Samsung Display Co. Yongin South Korea
-
A diffusion model is applied for defect image generation. Using an innovative method to automatically apply masks in diffusion sampling, the generated synthetic images are demonstrated to achieve high image quality and good defect diversity while preserving image fidelity outside the defective region when compared to the input normal image.
32.2 - Multi AI Approaches for Improving OLED Display Pattern Repair in Manufacturing Processes (9:20 AM - 9:40 AM)
-
Hong-bin Lim, Eun-chul Shin
Samsung Display Asan South Korea
-
To detect overlap areas between layers and defects in gap regions, the authors explore the use of deep learning (DL) techniques. By developing a DL auto-repair process based on accurate segmentation detection of repair targets, they establish a stable system by combining classification DL and detection DL.
32.3 - Heterogeneous Resource-Constrained Reinforcement Learning Photolithography Scheduler with Heterogeneous Graph Attention Network (9:40 AM - 10:00 AM)
-
Shuhui Qu, Kasra Yazdani, Janghwan Lee
Samsung Display American Lab San Jose CA US
-
An innovative framework employing a heterogeneous graph neural network-based rainbow algorithm is introduced to optimize scheduling in the dynamic photolithography process. A graph attention network-based architecture is implemented for deep representation learning. Reinforcement learning agents leverage the embeddings to prioritize products via a parameterized Q-function.