Learning View Priors for Single-view 3D Reconstruction

Abstract

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.

arXiv commentCVPR 2019. Project page: http://hiroharu-kato.com/projects_en/view_prior_learning.html

References in corpus (11)

• Adam: A Method for Stochastic Optimization
• Conditional Generative Adversarial Nets
• Deep Residual Learning for Image Recognition
• Spectral Normalization for Generative Adversarial Networks
• cGANs with Projection Discriminator
• Pixel2Mesh: Generating 3D Mesh Models from Single RGB Images
• Matryoshka Networks: Predicting 3D Geometry via Nested Shape Layers
• A Point Set Generation Network for 3D Object Reconstruction from a Single Image
• Rethinking Reprojection: Closing the Loop for Pose-aware ShapeReconstruction from a Single Image
• Iterative Instance Segmentation
• Learning Shape Priors for Single-View 3D Completion and Reconstruction