Affiliation:
1. Johns Hopkins University
Abstract
Abstract
Image-based modeling refers to the process of constructing a three-dimensional (3D) model based on a set of two dimensional (2D) images. With the development of computer vision and related technologies, image-based modeling has shown considerable potential in recent years. Traditional image-based modeling techniques, such as photogrammetry, structure from motion (SfM), and multi-view stereo (MVS) have some flaws and limitations that prevent widespread application of the technology such as a high demand of high-quality images, specific angular perspectives, and limited replication similarity. Artificial intelligence (AI), particularly advancements in deep learning, provides more chances for image-based modeling. The ability of deep learning algorithms to learn from vast amounts of data, generalize to new scenarios, and adapt to the complexities of the real world has unlocked new possibilities for generating more accurate and detailed 3D models from 2D images. This paper will introduce traditional image-based modeling techniques and discuss the potential of deep learning methods in image-based modeling technique and its future development prospects.
Publisher
Research Square Platform LLC