Learning View Priors for Single-view 3D Reconstruction
This addresses the challenge of generating plausible 3D reconstructions from limited views for applications in computer vision and graphics, representing an incremental improvement over existing methods.
The paper tackles the problem of ambiguous 3D shape reconstruction from single or few views by training a discriminator to learn view priors, enabling shapes that look reasonable from any viewpoint, and it outperforms state-of-the-art methods on synthetic and natural image datasets.
There is some ambiguity in the 3D shape of an object when the number of observed views is small. Because of this ambiguity, although a 3D object reconstructor can be trained using a single view or a few views per object, reconstructed shapes only fit the observed views and appear incorrect from the unobserved viewpoints. To reconstruct shapes that look reasonable from any viewpoint, we propose to train a discriminator that learns prior knowledge regarding possible views. The discriminator is trained to distinguish the reconstructed views of the observed viewpoints from those of the unobserved viewpoints. The reconstructor is trained to correct unobserved views by fooling the discriminator. Our method outperforms current state-of-the-art methods on both synthetic and natural image datasets; this validates the effectiveness of our method.