Detection of Geometric Elements in Axonometric Projection Images Using Computer Vision Techniques

Apaza Condori J.P.; Villanueva E. · IEEE Andescon, ANDESCON 2024 - Proceedings (2024) · DOI: 10.1109/andescon61840.2024.10755621

Resumen

The development of three-dimensional geometry problem-solving skills is important for engineering students and professionals. Acquiring these skills through traditional methods often presents challenges for students, mainly because the teaching material is presented in plain images with which students cannot interact to gain greater understanding. This paper describes the development of a computational tool designed to transform plain axonometric projection images into 3D models compatible with CAD design software. To achieve this, we have engineered deep learning models to accurately classify, locate, and identify geometric elements within axonometric images. Our experimental results highlight the enhanced performance of the Faster-RCNN model, which achieved an 81 % mean Average Precision (mAP) on test images. This tool can help engineering students to have a better learning experience, allowing them to interact with three-dimensional representations and significantly improve their spatial understanding.

Instituciones: Pontificia Universidad Católica del Perú