Development of 3D Modeling System from 2D Plane Figures using AR and AI

Abstract

Spatial figures are generally drawn in 2D such as in textbooks. While learning with real objects allows for intuitive understanding, learning only with textbooks requires a higher level of spatial ability. In this paper, an application was developed, that supports learners without enough spatial ability by using AR to visualize 2D figures in 3D, making it easier to grasp the image of the 2D to 3D transformation. The system generates 3D data from the features of a 2D plan view using OpenCV library and machine learning technology, and support learning with generic AR. Demonstration experiments was conducted with 67 and 24 users. In spatial figure problems, the group using AR tended to get higher accuracy than the group without AR in problems using the ratio of the line lengths and using rotations of 3D objects. The effectiveness of the developing system was verified from the questionnaire after the experiment.