Ponder: Point Cloud Pre-training via Neural Rendering
This addresses the problem of learning informative point cloud features for various 3D vision tasks, but it appears incremental as it builds on existing pre-training methods with a novel rendering approach.
The paper tackles self-supervised learning of point cloud representations by using differentiable neural rendering to train an encoder that renders realistic images from point clouds, achieving superior performance in downstream tasks like 3D detection and segmentation.
We propose a novel approach to self-supervised learning of point cloud representations by differentiable neural rendering. Motivated by the fact that informative point cloud features should be able to encode rich geometry and appearance cues and render realistic images, we train a point-cloud encoder within a devised point-based neural renderer by comparing the rendered images with real images on massive RGB-D data. The learned point-cloud encoder can be easily integrated into various downstream tasks, including not only high-level tasks like 3D detection and segmentation, but low-level tasks like 3D reconstruction and image synthesis. Extensive experiments on various tasks demonstrate the superiority of our approach compared to existing pre-training methods.