No window open for 24 hours.     

MENU
Conference

Speaker Info

Home  >   Conference   >   Speaker Info



  기조강연 2025.6.5 9:40 ~ 2025.6.5 10:20

[Special Speech] Robust Visual Intelligence for Manufacturing: Continual Learning and Realistic Noise Synthesis
YoungJoon Yoo,
Chief Research Officer, SNUAILAB
In this session, I will introduce two studies on AI-based visual recognition technology for manufacturing automaticization technology.The first announcement, introduces a new benchmark for detection performance evaluation performance evaluation in reality learning environment.This benchmark is expanding existing detection environment, and large-scale evaluation environment, and existing representative public data sets and existing representative public data sets.In particular, the new evaluation scenario that can measure the zero-shot generalization performance to measure the actual industrial environment that cant see more closely related to actual industrial environment.The announcement is scheduled to introduce insights and new base line models obtained through various experiments.The second announcement provides a new generation model for creating synthetic noise near real noise.This method can create noise integrated noise in various noise environments using only one no-clean image pairs of noise environments, such as traditional metadata or special equipment.In addition, the RefNo-clean pair can easily generate powerful image restore performance in real manufacturing environment and lightweight models in real manufacturing environment in real manufacturing environment.Both studies expectations that can be applied to actual manufacturing environment, and generalization will give people a personality and general manufacturing environment.