perfectism's blog

物来顺应,未来不迎,当时不杂,既往不恋

0%

Numpy 笔记

transpose

1
2
3
4
5
6
7
8
9
10
11
12
numpy.transpose(a, axes=None)
#修改张量的维度
>>> x = np.arange(4).reshape((2,2))
>>> x
array([[0, 1],
[2, 3]])
>>> np.transpose(x)
array([[0, 2],
[1, 3]])
>>> x = np.ones((1, 2, 3))
>>> np.transpose(x, (1, 0, 2)).shape
(2, 1, 3)
阅读全文 »

Notes of Machine Learning

Min Chen Yang, Hunan University, 2019.11

初识机器学习

从原始数据中提取模式的能力。

进入21世纪,纵观机器学习发展历程,研究热点可以简单总结为2000-2006年的流形学习、2006年-2011年的稀疏学习、2012年至今的深度学习 、未来迁移学习?

人工智能的真正挑战在于解决对人来说很容易执行,但很难形式化描述的任务。

常用的10大机器学习算法有:决策树、随机森林、逻辑回归、SVM、朴素贝叶斯、K最近邻算法、K均值算法、Adaboost算法、神经网络、马尔科夫

阅读全文 »

Note of ROS to Ubantu

terminal命令

echo $PATH: echo $PATH用于列出变量PATH的值,里面包含了已添加的目录

env:env命令是environment的缩写,用于列出所有的环境变量

export:单独使用export命令也可以像env列出所有的环境变量,不过export命令还有其他额外的功能

New Words

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
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
scalar 标量的
neuron 神经元
neural 神经的
vector 向量
parameter 参数
gradient descent 梯度下降
backpropagation 反向传播
sentiment 感情、情绪
embedded 嵌入的、植入的
trimming 整理、修剪、切除
syntactic parsing 语法分析
semantic analysis 语义分析
indicator 指示器,指示符
shallow 浅的
adversarial 对抗的
indent 缩进
contour 轮廓
approx=approxiate 大约的
hierarchy 层级,等级制度
Skeleton Joint 骨骼关节
Semantic Segmentation 语义分割
advent 来临
gaze 注视
row 行
colunm 列
extremum 极值
gray level 灰度
entropy 熵
stochastic 随机的,猜测的
crop 修剪,种植
shuffle 随机播放,打乱顺序
transpose 调换,转置
precisely 恰好地
dilation 扩张
co-adaptation 过拟合
syllabus 教学摘要,课程表
spatial 空间的
stochastic 随机的
bottleneck 瓶颈
geometry 几何结构
plain 平的,简单的,朴素的
Fine-tune 微调
spatial 空间的
sparse 稀疏的
backend 后端
attribute 属性
deque 双端队列
combat 反抗战斗
temporal,temporary 暂时的

Detection of Secondary Driving Tasks

MIT—AVT

王飞跃等的相关工作

1.End-to-End Driving Activities and Secondary Tasks Recognition Using Deep Convolutional Neural Network and Transfer Learning

  • Specifically, seven common driving activities are identified, which are normal driving, right mirror checking, rear mirror checking, left mirror checking, using in-vehicle video device, texting, and answering mobile phone. Among these, the first four activities are regarded as normal driving tasks, while the rest three are divided into distraction group.
  • using a Gaussian mixture model (GMM) to extract the driver region from the background
  • AlexNet:directly takes the processed RGB images as the input and outputs the identified label.
  • to reduce the training cost, the transfer learning mechanism is applied
  • An average of 79% detection accurac
阅读全文 »

CNN

Convolutional Neural Networks

Foundation

卷积层(Convolutional layer)

作用:对输入的数据进行特征提取,可对应多个卷积核,一般来说经过卷积层后,尺寸变小,厚度变厚,越往后面感受野变大,既降尺寸,又降厚度,参数需要学习

卷积核和图像一样有通道数,卷积核滑动步长(stride),每个卷积核对应一个局部特征(梯度特征)

为了让输入图像的全部像素都能被滑动窗口捕获,使用same方式在输入图像最外层加上指定层数(padding)全为0的像素边界,使用valid方式不做像素填充。经过卷积后输出图像的宽度与高度参数如下:

阅读全文 »

计算机视觉:算法与应用 笔记

1 图像形成

1.1.1 几何基元与变换

几何基元:2D点、2D直线、2D圆锥曲线、3D点、3D平面、3D直线、3D二次曲面

齐次坐标(投影坐标):用于投影几何里的坐标系统,给定欧氏平面上的一点 (x, y),对任意非零实数 Z,三元组 (xZ, yZ, Z) 即称之为该点的齐次坐标

笛卡尔坐标与齐次坐标的关系

  • 投影平面上的任何点都可以表示成一三元组 (X, Y, Z),称之为该点的齐次坐标投影坐标,其中 X、Y 及 Z 不全为 0

  • 当 Z 为 0,则该点表示一无穷远点

  • 三元组 (0, 0, 0) 不表示任何点。原点表示为 (0, 0, 1)

阅读全文 »

C++学习笔记

闵晨阳

1 C++初识

1.1 第一个C++程序

编写一个C++程序总共分为4个步骤

  • 创建项目
  • 创建文件
  • 编写代码
  • 运行程序

1.1.1 创建项目

​ Visual Studio是我们用来编写C++程序的主要工具,我们先将它打开

1541383178746

1541384366413

1.1.2 创建文件

右键源文件,选择添加->新建项

1541383817248

给C++文件起个名称,然后点击添加即可。

阅读全文 »

Note of Opencv to Python

Opencv 在Pycharm中的配置

1
2
3
pip install opencv-python
pip install opencv-contrib-python #扩展库
pip install pytesseract

新建python项目,注意解释器正确配置应该如下:

1571710989690

常见图像坐标系

Opencv:

Matlab:

阅读全文 »