Point-M2AE: Multi-scale Masked Autoencoders for Hierarchical Point Cloud Pre-training.
Renrui ZhangZiyu GuoPeng GaoRongyao FangBin ZhaoDong WangYu QiaoHongsheng LiPublished in: CoRR (2022)
Keyphrases
- point cloud
- multiscale
- surface reconstruction
- structure from motion
- coarse to fine
- restricted boltzmann machine
- normal vectors
- denoising
- feedforward neural networks
- point cloud data
- laser scanner
- point sets
- real world objects
- image processing
- stereo camera
- triangular mesh
- deep belief networks
- hough transform
- viewpoint
- high dimensional
- missing data
- image registration
- training data
- icp algorithm
- computer vision