PointInfinity: Resolution-Invariant Point Diffusion Models
This addresses the challenge of efficient high-resolution point cloud generation for 3D vision applications, representing an incremental advance over existing methods like Point-E.
The authors tackled the problem of generating high-resolution point clouds efficiently by introducing PointInfinity, a diffusion model with a resolution-invariant latent representation, enabling training on low-resolution data and inference at higher resolutions, which improved fidelity and scaled up to 131k points with state-of-the-art quality.
We present PointInfinity, an efficient family of point cloud diffusion models. Our core idea is to use a transformer-based architecture with a fixed-size, resolution-invariant latent representation. This enables efficient training with low-resolution point clouds, while allowing high-resolution point clouds to be generated during inference. More importantly, we show that scaling the test-time resolution beyond the training resolution improves the fidelity of generated point clouds and surfaces. We analyze this phenomenon and draw a link to classifier-free guidance commonly used in diffusion models, demonstrating that both allow trading off fidelity and variability during inference. Experiments on CO3D show that PointInfinity can efficiently generate high-resolution point clouds (up to 131k points, 31 times more than Point-E) with state-of-the-art quality.