In recent years, neural network-based methods have been widely used in camera-based 3D reconstruction.
But in most cases, 3D reconstruction still requires hundreds of camera perspectives to complete.
At the same time, while traditional photometric methods can calculate high-precision reconstruction results (even for objects with no surface texture), these methods are usually only effective under controlled laboratory conditions.
, According to foreign media reports, Daniel Cremers, professor of Computer Vision and Artificial Intelligence at Technical University Munich, head of the Munich Machine Learning Center (MCML) and director of the Munich Data Science Institute (MDSI), and his team jointly developed a method to achieve 3D reconstruction using only two camera perspectives.
, Image source: Technical University Munich, This method combines a neural network of the surface with an accurate model of the lighting process that takes into account light absorption and the distance between the object and the light source.
The brightness in the image can be used to determine the angle and distance of the surface relative to the light source.
, Return to the first electric network home page>,.
