상세정보
학술대회
Magnifier Network for Ceramic Z-directional Height Estimation with a Single Image
- 발행일
- 202010
- 출처
- International Conference on Information and Communication Technology Convergence (ICTC) 2020, pp.1199-1202
- 협약과제
- 20PS1400, 세라믹산업 제조혁신을 위한 클라우드 기반 빅데이터 플랫폼 개발, 지수영
- 초록
- Due to the AI development, productivity of manufactures has been increased. This paper introduces a method that detects three dimensional size defects in ceramic products using a monocular depth estimation method. Conventional depth estimation is weak to deal with two-sided biased data. This paper addresses a brand new magnifier network in order to magnify important parts among an input with various ranges. Deep neural networks have considered that they are weak to learn partial area. However, the proposed method overcomes this weak point by applying to a proposed magnifier method. In this paper, we present that the proposed method increases the performance in terms of qualitative and quantitative aspects.
- KSP 제안 키워드
- Ceramic products, Deep neural network(DNN), Estimation method, Monocular depth estimation, Partial area, Single image, Three dimensional(3D), Three dimensional size, height estimation