Self-calibrating Deep Photometric Stereo Networks
This work addresses the challenge of 3D reconstruction from images under unknown lighting for computer vision applications, representing a novel approach but with incremental advancements in deep learning architectures.
The paper tackles the problem of uncalibrated photometric stereo for non-Lambertian scenes by proposing a deep learning method that estimates shape and light directions without relying on specific reflectance or light assumptions, achieving significant performance improvements over previous methods.
This paper proposes an uncalibrated photometric stereo method for non-Lambertian scenes based on deep learning. Unlike previous approaches that heavily rely on assumptions of specific reflectances and light source distributions, our method is able to determine both shape and light directions of a scene with unknown arbitrary reflectances observed under unknown varying light directions. To achieve this goal, we propose a two-stage deep learning architecture, called SDPS-Net, which can effectively take advantage of intermediate supervision, resulting in reduced learning difficulty compared to a single-stage model. Experiments on both synthetic and real datasets show that our proposed approach significantly outperforms previous uncalibrated photometric stereo methods.