DeepShadow: Neural Shape from Shadow
This addresses the problem of 3D shape recovery in computer vision by leveraging shadows, which were previously considered disturbances, offering a novel approach for applications like robotics or graphics.
The paper tackles 3D reconstruction from photometric stereo shadow maps by proposing DeepShadow, a one-shot method that recovers depth maps and surface normals using shadows as a learning signal, showing it can outperform shape-from-shading in some cases.
This paper presents DeepShadow, a one-shot method for recovering the depth map and surface normals from photometric stereo shadow maps. Previous works that try to recover the surface normals from photometric stereo images treat cast shadows as a disturbance. We show that the self and cast shadows not only do not disturb 3D reconstruction, but can be used alone, as a strong learning signal, to recover the depth map and surface normals. We demonstrate that 3D reconstruction from shadows can even outperform shape-from-shading in certain cases. To the best of our knowledge, our method is the first to reconstruct 3D shape-from-shadows using neural networks. The method does not require any pre-training or expensive labeled data, and is optimized during inference time.