Discrete Point Flow Networks for Efficient Point Cloud Generation
arXiv:2007.10170
Abstract
Generative models have proven effective at modeling 3D shapes and their statistical variations. In this paper we investigate their application to point clouds, a 3D shape representation widely used in computer vision for which, however, only few generative models have yet been proposed. We introduce a latent variable model that builds on normalizing flows with affine coupling layers to generate 3D point clouds of an arbitrary size given a latent shape representation. To evaluate its benefits for shape modeling we apply this model for generation, autoencoding, and single-view shape reconstruction tasks. We improve over recent GAN-based models in terms of most metrics that assess generation and autoencoding. Compared to recent work based on continuous flows, our model offers a significant speedup in both training and inference times for similar or better performance. For single-view shape reconstruction we also obtain results on par with state-of-the-art voxel, point cloud, and mesh-based methods.
arXiv commentIn ECCV'20
References in corpus (12)
- Auto-Encoding Variational Bayes
- Decoupled Weight Decay Regularization
- Stochastic Backpropagation and Approximate Inference in Deep Generative Models
- On the Convergence of Adam and Beyond
- Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling
- Density estimation using Real NVP
- Learning Representations and Generative Models for 3D Point Clouds
- Variational Lossy Autoencoder
- SPLATNet: Sparse Lattice Networks for Point Cloud Processing
- 3D-LMNet: Latent Embedding Matching for Accurate and Diverse 3D Point Cloud Reconstruction from a Single Image
- Learning to Generate and Reconstruct 3D Meshes with only 2D Supervision
- Adaptive Density Estimation for Generative Models