Tensorflow：numpy基础

Tensorflow：numpy基础

tensorflow的运算基于numpy，numpy是矩阵数学运算的库。

1. numpy向量和矩阵乘法

   1.1 向量点乘

   >>> import numpy as np
   >>> a = np.array([2,0])
   >>> b=np.array([1,1])
   >>> np.dot(a,b)
   2
   

   1.2 矩阵乘法

   >>> import numpy as np
   >>> a = np.array([[1,2],[2,2],[3,4]])
   >>> b = np.array([[3,4,5,6],[4,5,6,7]])
   >>> print(a.shape, b.shape)
   (3, 2) (2, 4)
   >>> c = a.matmul(a,b)
   >>> print(c.shape)
   (3, 4)
   

2. opencv和numpy

   2.1 opencv读取的数据和numpy可以无缝转换

   >>> import cv2
   >>> img = cv2.imread(r"G:\model\target\target\380740c40a2f48bc942093c3e3b0b2e268.png")
   >>> print(img.shape)
   (130, 320, 3)
   

   2.2 shape依次为hwc

   h=130, w=320, c=3，c是channel通道，依次是b，g，r，取值都是0-255。

   2.3 numpy的基础运算，传播（broadcast）

   >>> a = np.array([[2,4,6,7],[1,3,5,6],[4,4,2,9]])
   >>> print(a.shape)
   (3, 4)
   >>> print(a)
   [[2 4 6 7]
    [1 3 5 6]
    [4 4 2 9]]
   >>> b=a/2
   >>> print(b)
   [[1.  2.  3.  3.5]
    [0.5 1.5 2.5 3. ]
    [2.  2.  1.  4.5]]
   >>> c = np.array([2,1,1,2])
   >>> print(a*c)
   [[ 4  4  6 14]
    [ 2  3  5 12]
    [ 8  4  2 18]]
   >>> d=np.array([[2,1,1,2],[0.5,1,1,0.5],[2,1,1,2]])
   >>> print(a*d)
   [[ 4.   4.   6.  14. ]
    [ 0.5  3.   5.   3. ]
    [ 8.   4.   2.  18. ]]
   

   简记：

   加减乘除一个数，所有元素都加减乘除

   加减乘除一个和最后一个维度相同的数组，最后一个维度对应位置进行加减乘除

   加减乘除一个和自己维度相同的矩阵，对应位置进行加减乘除

   2.4 numpy的切片

   >>> a = np.array([[2,4,6,7],[1,3,5,6],[4,4,2,9]])
   >>> print(a.shape)
   (3, 4)
   >>> print(a)
   [[2 4 6 7]
    [1 3 5 6]
    [4 4 2 9]]
   >>> print(a[1:,1:])
   [[3 5 6]
    [4 2 9]]
   >>> print(a[0:2,1:2])
   [[4]
    [3]]
   
   
   

注意：切片的区间前闭后开