Point-M2AE: Multi-scale Masked Autoencoders for Hierarchical Point Cloud Pre-training.
Renrui ZhangZiyu GuoPeng GaoRongyao FangBin ZhaoDong WangYu QiaoHongsheng LiPublished in: NeurIPS (2022)
Keyphrases
- point cloud
- multiscale
- coarse to fine
- restricted boltzmann machine
- normal vectors
- surface reconstruction
- structure from motion
- point sets
- stereo camera
- laser scanner
- point cloud data
- training set
- denoising
- supervised learning
- real world objects
- feedforward neural networks
- triangular mesh
- computer vision
- training samples
- sample points
- image segmentation
- image processing
- deep belief networks