Development of a Specific Shape Recognition Algorithm in 2D Drawing of TV or Monitor Cover and Chassis Using Cascade R-CNN
Object Detection
Objects' Orientation Estimation
As the replacement cycle of products is shortened due to various demands from consumers, the demand for mold manufacturing is increasing and the delivery period is getting shorter. The disadvantage of the mold design automation system developed to date is that it relies solely on the competence of the mold designer. Although CAD developers have released many shape analysis tools to reduce dependence on workers' capabilities, they cannot apply them to current businesses. Thus, in this study, cascade R-CNN was used to recognize certain features in TV covers and chassis 2D drawings. As a result of recognition, the shape part is displayed in the form of a boundary box, and the type of feature, the location from the center of the drawing, the degree of rotation (90 degree intervals), and the size of the boundary box by the width and width.
Donation: Samsung Electronics Global Technology Center
Term : 2019. 04. 01 ~ 2019. 12. 31.
Expenses : 60,000,000 WON
Name: Bak Seongho, Ph.D Candidate
E-mail: bakseongho@gm.gist.ac.kr