3D Vision Notes | Point Clouds Processing 01 | Introduction
Contents
1. Introduction and Basic Algorithm
Classical Algorithms vs Deep Learning
| - | Object Classification | Object Registration | Object Detection |
|---|---|---|---|
| Classical | Keypoint Detection Keypoint Description SVM |
Nearest Neighbor Search Iterative Closest Point(ICP) |
Background Removal Clustering Classification |
| Deep Learning | Data Collection Data Labeling Train a Network |
Data Collection Data Labeling Train a Network |
Data Collection Data Labeling Train a Network |
Note Lists:
- Introduction and Basics Algorithms
- PCA and Kernel PCA
- Smoothing, Filtering and Downsampling
- Nearest Neighbor Problem
- Binary Search Tree
- KD-Tree
- Octree
- Clustering
- K-Means
- Gaussian Mixture Model (GMM)
- Expectation-Maximization (EM)
- Spectral Clustering
- Model Fitting
- Meanshift & dbscan
- Robust Least Square
- Hough Transform
- RANSAC
- Deep Learning on Point Cloud
- PointNet & PointNet++
- GCN
- Supplementary
- 3D Object Detection
- RCNN, FastRCNN, FasterRCNN, SSD
- VoxelNet, PointPillars
- PointRCNN
- Frustum PointNet, PointPainting
- 3D Feature Detection
- harris 2d, 3d, 6d
- ISS: Intrinsic Shape Signatures
- USIP
- SO-Net
- 3D Feature Description
- PFH & FPFH
- SHOT
- 3D Match & Pe
- Registration (!!!!)
- ICP:
- NDT: Normal Distribution Transform
- Registration by RANSAC with feature detection, description, matching
深度学习：简单却不太容易解释; 要理解传统方法