Pointnet++个人笔记

前言

Pointnet提取的全局特征能够很好地完成分类任务，由于网络将所有的点最大池化为了一个全局特征，因此局部点与点之间的联系并没有被网络学习到，导致网络的输出缺乏点云的局部结构特征，因此PointNet对于场景的分割效果十分一般。在点云分类和物体的Part Segmentation中，这样的问题可以通过中心化物体的坐标轴部分地解决，但在场景分割中很难去解决。

前言

Pointnet开创性地将深度学习直接用于三维点云任务。由于点云数据的无序性，无法直接对原始点云使用卷积等操作。Pointnet提出对称函数来解决点的无序性问题，设计了能够进行分类和分割任务的网络结构，本文结合源码与个人的理解对于T-net网络和对称函数进行分析。

Since 2017

Sort out papers on point cloud research from 2017, and add keywords for the research area. For anyone who wants to do research about 3D point cloud, thanks for the Open source community .If you find the awesome paper/code/dataset or have some suggestions, please contact linhua2017@ia.ac.cn.

Few prior works study deep learning on point sets. PointNet by Qi et al. is a pioneer in this direction. However, by design PointNet does not capture local structures induced by the metric space points live in, limiting its ability to recognize fine-grained patterns and generalizability to complex scenes. In this work, they introduce a hierarchical neural network that applies PointNet recursively on a nested partitioning of the input point set. By exploiting metric space distances, our network is able to learn local features with increasing contextual scales. With further observation that point sets are usually sampled with varying densities, which results in greatly decreased performance for networks trained on uniform densities, we propose novel set learning layers to adaptively combine features from multiple scales. Experiments show that our network called PointNet++ is able to learn deep point set features efficiently and robustly. In particular, results significantly better than state-of-the-art have been obtained on challenging benchmarks of 3D point clouds.

1. Architecture

数据科学家技能地铁图

逻辑回归代价函数的求导详细推理过程

Point cloud viewer

Dependencies

1
2
3

pip install numpy --user
pip install mayavi --user
pip install PyQt5 --user<!---more--->

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

import numpy as np
import mayavi.mlab #Python for 3D plotting
"""
3D point cloud visualization
"""
def viz_mayavi(points, vals="distance"):
    x = points[:, 0]  # x 
    y = points[:, 1]  # y 
    z = points[:, 2]  # z 
    # r = lidar[:, 3]  # Confidence value
    d = np.sqrt(x ** 2 + y ** 2)  # Distance from image to touch sensor

    if vals == "height":
        col = z
    else:
        col = d

    fig = mayavi.mlab.figure(bgcolor=(0, 0, 0), size=(640, 360))
    mayavi.mlab.points3d(x, y, z,
                         col,  # Color value
                         mode="point",
                         colormap='spectral',  # 'bone', 'copper', 'gnuplot'
                         # color=(0, 1, 0),   # Use fixed colors
                         figure=fig,
                         )
    mayavi.mlab.show()


points = np.loadtxt('data/6.xyz')# Select data
viz_mayavi(points)

Demo screenshot

零基础学SVM—Support Vector Machine(一)

如果你是一名模式识别专业的研究生，又或者你是机器学习爱好者，SVM是一个你避不开的问题。如果你只是有一堆数据需要SVM帮你处理一下，那么无论是Matlab的SVM工具箱，LIBSVM还是python框架下的SciKit Learn都可以提供方便快捷的解决方案。但如果你要追求的不仅仅是会用，还希望挑战一下“理解”这个层次，那么你就需要面对一大堆你可能从来没听过的名词，比如：非线性约束条件下的最优化、KKT条件、拉格朗日对偶、最大间隔、最优下界、核函数等等。这些名词往往会跟随一大堆天书一般的公式。如果你稍微有一点数学基础，那么单个公式你可能看得明白，但是怎么从一个公式跳到另一个公式就让人十分费解了，而最让人糊涂的其实并不是公式推导，而是如果把这些公式和你脑子里空间构想联系起来。

基于sk-learn_多种模型对Boston房价预测对比

使用sk-learn进行房价预测的完整过程。