PRED: Pre-training via Semantic Rendering on LiDAR Point Clouds.
Hao YangHaiyang WangDi DaiLiwei WangPublished in: NeurIPS (2023)
Keyphrases
- point cloud
- photorealistic
- surface reconstruction
- structure from motion
- point sets
- lidar data
- stereo camera
- implicit surfaces
- point cloud data
- laser scanner
- triangular mesh
- ray tracing
- icp algorithm
- computer graphics
- high fidelity
- multi view reconstruction
- supervised learning
- viewpoint
- training set
- normal vectors
- high quality
- neural network
- nearest neighbor
- feature space
- multi view stereo