Generative Models for 3D Point Clouds
This work addresses the problem of enhancing 3D point cloud generation for applications in computer vision and graphics, but it appears incremental as it builds on existing methods without claiming major breakthroughs.
The paper tackled improving point cloud generative models by experimenting with transformer encoders, latent-space flow models, and autoregressive decoders, resulting in analysis and comparison of generation and reconstruction performance across various object types.
Point clouds are rich geometric data structures, where their three dimensional structure offers an excellent domain for understanding the representation learning and generative modeling in 3D space. In this work, we aim to improve the performance of point cloud latent-space generative models by experimenting with transformer encoders, latent-space flow models, and autoregressive decoders. We analyze and compare both generation and reconstruction performance of these models on various object types.