Title: Intelligent Perception in Amazon Consumer Robotics
GenAI in Semiconductor Manufacturing
Date：2024/12/27 14:20-15:30
Location：CSIE R103
Speakers: Prof. Min Sun,NTHU
Prof. Prof. Yuan-Fu Yang, NYCU
Host: Prof. Chun-Yi Lee
Topic: Intelligent Perception in Amazon Consumer Robotics
Abstract: In this talk, we will introduce Astro, Amazon's first consumer robot. We will share the journey of developing its intelligent perception capabilities and provide insights into our ongoing research efforts.
Bio: Professor Sun Min is a faculty member in the Department of Electrical Engineering at National Tsing Hua University, specializing in computer vision, deep learning, and reinforcement learning. In addition to his academic role, he serves as the Applied Scientist Lead at Amazon Consumer Robotics in Taiwan. Recognized for his exceptional work, Professor Sun has received numerous prestigious awards, including the Young Scholar Innovation Award from the Foundation for the Advancement of Outstanding Scholarship and the Wu Ta-You Memorial Award from the National Science Council. Notably, he won the Best Paper Award at the Computer Vision, Graphics, and Image Processing (CVGIP) conference for three consecutive years. From 2018 to 2022, Professor Sun served as Chief AI Scientist at Appier, where he played a crucial role in driving the company's successful public listing on the Tokyo Stock Exchange. His contributions continue to advance the frontiers of AI and robotics research and development.
Topic: GenAI in Semiconductor Manufacturing
Abstract: The semiconductor industry is at the forefront of technological innovation, driven by the demand for faster, more efficient chip designs. This tall will delve into three pivotal advancements in generative AI that enhance productivity and precision in this highly competitive field. First, HotspotFusion addresses Chemical Mechanical Polishing (CMP) hotspot detection, a key factor in improving New Tape Out (NTO) cycle times. By utilizing pattern density data from Graphic Design System (GDS) in a generative AI model, HotspotFusion enables proactive CMP hotspot detection early in the design phase, minimizing delays and optimizing process efficiency. Second, we present the Implicit Cross-Domain Diffusion Model (ICDDM), a weakly supervised approach to defect detection without pixel-wise annotated data. ICDDM leverages generative diffusion in low-dimensional latent spaces to identify defect patterns, with additional acceleration achieved through knowledge distillation. This model reduces inference time while maintaining performance, ensuring rapid defect detection and enhanced yield. Lastly, the Implicit Knowledge Distillation Diffusion Transformer (IKDDiT) offers a cutting-edge approach to overlay map generation in photolithography. Integrating text-to-image diffusion models and contrastive learning, IKDDiT enhances overlay alignment and reduces errors across photolithography layers, resulting in fewer reworks and improved productivity. Through these generative AI advancements, semiconductor manufacturing stands to gain unprecedented levels of precision, efficiency, and competitive advantage.
Bio: YuanFu Yang is an assistant professor at National Yang Ming Chiao Tung University with over a decade of experience in the IT department at TSMC, where he played a pivotal role in driving the company’s digital transformation. Yang led significant AI projects, including tGenie (TSMC's LLM/LMM), tAIOps, DLDR(Deep Learning Defect Review), and the Intelligent Scheduling System, all of which have advanced TSMC’s operational efficiency and innovation capabilities. His academic research focuses on computer vision and GenAI, with an emphasis on enhancing human creativity and advancing smart manufacturing. Yang’s work spans generative models, 3D representation, quantum deep learning, and large-scale multimodal models. In smart manufacturing, he is dedicated to optimizing production processes, increasing efficiency, and reducing costs through AI-driven technologies. His projects include the development of intelligent scheduling, process control, and logistics systems that dynamically adapt and optimize manufacturing environments. Through these contributions, Yang aims to drive industrial upgrades, foster technological innovation, and expand AI applications across diverse sectors.
[2024-12-27] Prof. Min Sun/Prof. Prof. Yuan-Fu Yang, NTHU/NYCU,