numpy的使用教程

数组的定义

import math

import numpy as np
import random
from timeit import timeit

# 定义
arr = np.array([1, 2])
print(arr)
'
运行
运行

随机数组的生成

# rand
rand = np.random.rand(100, 100)
print(rand)

数组的形状

# 判断形状
print(rand.size)
print(rand.shape)

np的意义

np=1.1735659999999999, python=20.3648931

def np_log():
    rand = np.random.rand(1000, 10000)
    x = np.log(rand)

def python_log():
    rand = []
    for i in range(1000):
        cur = []
        for j in range(10000):
            cur.append(random.random())
        rand.append(cur)

    for cur in rand:
        # print(cur)
        cur = [math.log(x) for x in cur]

# python_log()

# 定时
np_t = timeit('np_log()', 'from __main__ import np_log', number=10)
py_t = timeit('python_log()', 'from __main__ import python_log', number=10)
print(f"np={np_t}, python={py_t}")
def np_log():
    rand = np.random.rand(1000, 1000)
    x = np.log(rand)

def python_log():
    rand = []
    for i in range(1000):
        cur = []
        for j in range(1000):
            cur.append(random.random())
        rand.append(cur)

    for cur in rand:
        print(cur)
        cur = [math.log(x) for x in cur]

# python_log()

# 定时
np_t = timeit('np_log()', 'from __main__ import np_log', number=10)
py_t = timeit('python_log()', 'from __main__ import python_log', number=10)
print(f"np={np_t}, python={py_t}")

视频地址

https://www.bilibili.com/video/BV1Wy4y1h7ii/?spm_id_from=333.337.search-card.all.click&vd_source=82dc2acb60a90c43a2ac0d4023a2cd34

定义数组的几种方式

import numpy as np

# 定义数组
a = np.array([1,3,4,5])
print(f"define arr {a}")

z_3_2 = np.zeros((3, 2))  # 3 第一位 也就是多少行 3行2列
print(f"z_3_2 is {z_3_2}")

ar = np.arange(2, 8) # 类似于 range
print(f"a range is {ar}")

ls = np.linspace(0, 1, 5)
print(f"linspace is {ls}")

rad = np.random.rand(2, 4)
print(f"rand is {rad}")
"""
define arr [1 3 4 5]
z_3_2 is [[0. 0.]
 [0. 0.]
 [0. 0.]]
a range is [2 3 4 5 6 7]
linspace is [0.   0.25 0.5  0.75 1.  ]
rand is [[0.08905149 0.89795301 0.665294   0.93021576]
 [0.84545149 0.70275937 0.4369173  0.17938872]]
 """

1. array, 由python数组转过来
2. zeros ones 注意传过来的是一个元组
3. rand 注意传过来的是两个参数
4. arange
5. linspace

数据的类型

1. 默认是np.float64

2. 查看方式 a.dtype

3. 可以在定义的时候指定， np.zeros((4, 2), dtype=np.int32)

4. 类型的转化 b = a.astype(int)

5. 常用类型

   

   基本运算

   1. 同位置 ±*/

   2. dot 点乘

   3. np.sin cos, log power(a, 3), sqrt 每个元素

   4. arr * 常数 广播

   5. @ 矩阵乘法

      a = np.array([[1, 2],
                    [3, 4]])
      b = np.array([[2, 0],
                    [0, 2]])
      print(f"a={a} \n b={b}")
      print(f"np.dot([1, 2], [1, 2]) {np.dot([1, 2], [1, 2])}")
      print(f"* multiply {a * b}") # 等同于 np.multiply(), 逐个元素运算
      print(f"@  matmul {b @ a}") # 等同于 np.matmul()   矩阵乘法
      

      

特殊的计算方式

1. 乘以一个常数 — 广播

2. 一个行矩阵可以和一个列矩阵直接相乘

   

   

统计类的运算

1. np.max

2. np.argmax 索引

3. np.sum

4. np.mean

5. np.median

6. a.var() 方差

7. a.std() 标准差

多维矩阵的统计 axis

1. axis = 0，表示沿着行的方向计算

   

2. axis = 1， 沿着列的方向计算

   

获取某个元素

a = np.array([[1, 2, 3,],
			[2, 5, 6]])
a[0, 1] # 0 行 1列的元素

筛选元素

a = np.arange(10)  # array([0, 1, 2 ....9])
a[a<3]   # array([0, 1, 2])
a[ (a>3) & (a % 2 == 0)]  # array([4, 6, 8])  大于3而且是偶数

切片语法

第0行的前面两列元素。a[0, 0:2]

第一行 a[1, :] 或者 a[1]

切片 a[::-1]

numpy 处理图片

from PIL import Image
import numpy as np

i = Image.open('dog.jpg')
im = np.array(i)
# print(im)
print(im.shape)
print(im[100, 100])

im_r = im[:, :, 0]
# Image.fromarray(im_r).show() # 图片的显示

im_g = im[:, :, 1]
im_b = im[:, :, 2]
im_big = np.concatenate((im_r, im_g, im_b), axis=1)  # 图片的拼接
# Image.fromarray(im_big).show()

im_blend = im_r * 0.4 + im_g * 0.6 # 图像的混合
im_blend = im_blend.astype(np.uint8)
# Image.fromarray(im_blend).show()

im_downsample = im[::2, ::2, :] # 降采样
# Image.fromarray(im_downsample).show()

im_flipped = im[::-1, :, :]  # 翻转
Image.fromarray(im_flipped).show()

声音采集的例子

def get_wave_to_time(source_file_name, wave_one_pip_sample_count = 3000, should_plt = False, head_trim_time = 3, threhold = 0.07):
    time, signal, fps = play_audio_with_py_av_lib(source_file_name, should_plt, threhold)
    len = time.size

    samples_exceeding_threshold = np.where(signal > threhold)[0]
    filter_big_gap = np.diff(samples_exceeding_threshold) > wave_one_pip_sample_count
    filter_big_gap = np.insert(filter_big_gap, 0, True)
    samples_exceeding_threshold = samples_exceeding_threshold[filter_big_gap]
    # samples_exceeding_threshold = samples_exceeding_threshold / fps

    return time, signal, fps, time[samples_exceeding_threshold]

opencv 的例子

怎么应用：
https://stackoverflow.com/questions/47501249/implement-opencv-method-warpperspective
todo

import cv2
import numpy as np

img = cv2.imread("cards.jpg")

width,height = 250,350
pts1 = np.float32([[111,219],[287,188],[154,482],[352,440]])
pts2 = np.float32([[0,0],[width,0],[0,height],[width,height]])


def using_opencv():
    matrix = cv2.getPerspectiveTransform(pts1, pts2)
    print(matrix)
    imgOutput = cv2.warpPerspective(img, matrix, (width, height))
    # cv2.imshow("Image", img)
    # cv2.imshow("Output", imgOutput)
    # cv2.waitKey(0)



# https://blog.csdn.net/stf1065716904/article/details/92795238   根本的做法
# ndarry的  基本使用
def caculate(src, dst):
    assert src.shape[0] == dst.shape[0] and src.shape[0] >= 4

    nums = src.shape[0]
    A = np.zeros((2 * nums, 8))  # A*warpMatrix=B
    B = np.zeros((2 * nums, 1))
    for i in range(0, nums):
        A_i = src[i, :]
        B_i = dst[i, :]
        A[2 * i, :] = [A_i[0], A_i[1], 1,
                       0, 0, 0,
                       -A_i[0] * B_i[0], -A_i[1] * B_i[0]]
        B[2 * i] = B_i[0]

        A[2 * i + 1, :] = [0, 0, 0,
                           A_i[0], A_i[1], 1,
                           -A_i[0] * B_i[1], -A_i[1] * B_i[1]]
        B[2 * i + 1] = B_i[1]

    A = np.mat(A)
    warpMatrix = A.I * B  # 求出a_11, a_12, a_13, a_21, a_22, a_23, a_31, a_32

    # 之后为结果的后处理
    aa = np.array(warpMatrix)
    # warpMatrix = aa.T[0]
    warpMatrix = aa.reshape(-1)
    warpMatrix = np.insert(warpMatrix, warpMatrix.shape[0], values=1.0, axis=0)  # 插入a_33 = 1
    warpMatrix = warpMatrix.reshape((3, 3))
    print(warpMatrix)
    return warpMatrix

using_opencv()
caculate(pts1, pts2)

[[ 1.51412424e+00 -2.47556434e-01 -1.13852931e+02]
[ 2.90344820e-01 1.64840930e+00 -3.93229912e+02]
[-4.56552320e-05 5.83742088e-04 1.00000000e+00]]
[[ 1.51412424e+00 -2.47556434e-01 -1.13852931e+02]
[ 2.90344820e-01 1.64840930e+00 -3.93229912e+02]
[-4.56552320e-05 5.83742088e-04 1.00000000e+00]]

numpy 书籍

https://www.labri.fr/perso/nrougier/from-python-to-numpy

从python 到numpy

python技巧

zen

import this

具体情况

fruit = [x for x in fruit if x.startswith("a")] # 列表解析式 在最开始 ，if 在最后面
# 7 字典的合并  ** 叫做解包
a = {"a":1, "b":2}
b = {"z":26, "y":25}
c = {**a, **b}  


for index, x in enumerate(reversed(fruit)) # # 6 逆向遍历

for index, x, in enumerate(sorted(fruit)) # 6.2 排序
for index, x in enumerate(fruit):     # 5 得到



x = "a" if a = 1 else "b"  # 8

# 9 解包
name = "San zhang"
name, xing = name.split()
a, b = tuple
a, b = range(2)

a, b = b, a #1
f"a is {a}" #2
fruit = ["apple", "pear", "banana"]
fruit =  [x.upper() for x in fruit]  # 4,列表生成式  列表解析式

yield 具体例子

def fibonacci(n):
	a = 0
	b = 1
	for _ in range(n):
		yield a  # 生成器，在for循环过程中得到一个值就可以返回回去
		# 产出， return需要都做完再返回， yield中途可以返回
		a, b = b, a+b
		
for i in fibonacci(10):
	print(i)

线上的练习环境

http://repl.it