写得很详细：https://www.cnblogs.com/kevingrace/p/5690241.html（已剪报）
月份：2018年12月
numpy部分用法
1. np.argsort() >>> import numpy as np >>> a=np.array([1,-1,4,5,0]) >>> index=np.argsort(a) #元素从小到大排序，只返回索引List（原始数组元素位置不改变） >>> print(index) [1 4 0 2 3]
2.np.bincount() #bin=最大值加1，每个bin给出了它的索引值在x中出现的次数 #eg： # 我们可以看到x中最大的数为7，因此bin的数量为8，那么它的索引值为0->7 x = np.array([0, 1, 1, 3, 2, 1, 7]) # 索引0出现了1次，索引1出现了3次......索引5出现了0次...... np.bincount(x) #因此，输出结果为：array([1, 3, 1, 1, 0, 0, 0, 1]) # 我们可以看到x中最大的数为7，因此bin的数量为8，那么它的索引值为0->7 x = np.array([7, 6, 2, 1, 4]) # 索引0出现了0次，索引1出现了1次......索引5出现了0次...... np.bincount(x) #输出结果为：array([0, 1, 1, 0, 1, 0, 1, 1]) ************************* # weight参数 w = np.array([0.3, 0.5, 0.2, 0.7, 1., -0.6]) # 我们可以看到x中最大的数为4，因此bin的数量为5，那么它的索引值为0->4 x = np.array([2, 1, 3, 4, 4, 3]) # 索引0 -> 0 #因为0索引出现次数为0 # 索引1 -> w[1] = 0.5 #1索引出现次数为1，且对应权值为0.5 # 索引2 -> w[0] = 0.3 # 索引3 -> w[2] + w[5] = 0.2 - 0.6 = -0.4 # 3索引出现次数为2，对应两处权值的和 # 索引4 -> w[3] + w[4] = 0.7 + 1 = 1.7 np.bincount(x, weights=w) # 因此，输出结果为：array([ 0. , 0.5, 0.3, -0.4, 1.7]) ref：https://blog.csdn.net/xlinsist/article/details/51346523
3.np.vstack()
垂直排列数组
"""
Examples
--------
>>> a = np.array([1, 2, 3])
>>> b = np.array([2, 3, 4])
>>> np.vstack((a,b))
array([[1, 2, 3],
[2, 3, 4]])
>>> a = np.array([[1], [2], [3]])
>>> b = np.array([[2], [3], [4]])
>>> np.vstack((a,b))
array([[1],
[2],
[3],
[2],
[3],
[4]])
"""
4.np.hstack()
水平排列数组
"""
Examples
--------
>>> a = np.array((1,2,3))
>>> b = np.array((2,3,4))
>>> np.hstack((a,b))
array([1, 2, 3, 2, 3, 4])
>>> a = np.array([[1],[2],[3]])
>>> b = np.array([[2],[3],[4]])
>>> np.hstack((a,b))
array([[1, 2],
[2, 3],
[3, 4]])
"""
5. np.linspace()
'''
返回均匀间隔的数组,[start,stop]
Examples
--------
>>> np.linspace(2.0, 3.0, num=5)
array([ 2. , 2.25, 2.5 , 2.75, 3. ])
>>> np.linspace(2.0, 3.0, num=5, endpoint=False) #不包括stop
array([ 2. , 2.2, 2.4, 2.6, 2.8])
>>> np.linspace(2.0, 3.0, num=5, retstep=True) #返回步长
(array([ 2. , 2.25, 2.5 , 2.75, 3. ]), 0.25)
'''
6.np.prop()
# Return the product of array elements over a given axis
'''
Examples
--------
By default, calculate the product of all elements:
>>> np.prod([1.,2.])
2.0
Even when the input array is two-dimensional:
>>> np.prod([[1.,2.],[3.,4.]])
24.0
But we can also specify the axis over which to multiply:
>>> np.prod([[1.,2.],[3.,4.]], axis=1)
array([ 2., 12.])
'''
7.np.random.binomial(n,p,size)
'''
二项分布，在size次实验的每次实验中，n个硬币中出现正面的硬币个数
>>> np.random.binomial(3,0.5,10)
array([2, 2, 2, 1, 2, 1, 1, 2, 2, 1])
>>> np.sum(np.random.binomial(3,0.5,10)==2)/10 #10次实验中出现2次硬币正面的概率
0.4
dropout中的使用：
>>> np.random.binomial(1,0.5,(10,10)) #X.shape=(10,10),保留50%的结点
array([[1, 1, 0, 1, 0, 1, 1, 0, 0, 1],
[0, 1, 0, 1, 1, 0, 1, 1, 1, 1],
[0, 0, 1, 1, 1, 1, 0, 1, 1, 0],
[1, 0, 1, 0, 0, 0, 1, 0, 1, 0],
[0, 1, 0, 0, 0, 1, 0, 0, 0, 0],
[1, 0, 1, 1, 0, 0, 1, 1, 1, 1],
[1, 1, 0, 0, 0, 1, 1, 0, 1, 1],
[0, 0, 1, 0, 0, 1, 1, 0, 1, 0],
[0, 0, 0, 1, 0, 1, 1, 0, 1, 1],
[1, 1, 1, 1, 0, 1, 0, 0, 0, 0]])
上述数组再乘以X即得到dropout后结果
'''
8.np.pad(array, pad_width, mode, **kwargs)
'''
数组的填充，可用于卷积的图像填充
>>> a=np.array([[1,2],[3,4]])
>>> np.pad(a,((2,3),(3,2)),'constant',constant_values=(8, 9))
# (2,3)表示第一维即按行填充，数组行的最前面填充2行全8，最后面填充3行全9；
# (3,2)表示第二维即按列填充，数组列的最前面填充3列全8，最后面填充2行全9；
# (8,9)表示前和后的填充数量
array([[8, 8, 8, 8, 8, 9, 9],
[8, 8, 8, 8, 8, 9, 9],
[8, 8, 8, 1, 2, 9, 9],
[8, 8, 8, 3, 4, 9, 9],
[8, 8, 8, 9, 9, 9, 9],
[8, 8, 8, 9, 9, 9, 9],
[8, 8, 8, 9, 9, 9, 9]])
'''
9.np.newaxis()创建新轴 ''' Example >>> a = np.array([3,1,5,9]) >>> print(a,a.shape) [3 1 5 9] (4,) >>> b=a[np.newaxis,:] >>> print(b,b.shape) [[3 1 5 9]] (1, 4) >>> c=a[:,np.newaxis] >>> print(c,c.shape) [[3] [1] [5] [9]] (4, 1) '''
密码保护：pressure
HOG
def：https://baike.baidu.com/item/HOG/9738560
ref：https://blog.csdn.net/zouxy09/article/details/7929348
paper：http://lear.inrialpes.fr/people/triggs/pubs/Dalal-cvpr05.pdf
code：参考cs231n
def hog_feature(im):
"""Compute Histogram of Gradient (HOG) feature for an image
Modified from skimage.feature.hog
http://scikit-image.org/docs/dev/api/skimage.feature.html#hog
Reference:
Histograms of Oriented Gradients for Human Detection
Navneet Dalal and Bill Triggs, CVPR 2005
Parameters:
im : an input grayscale or rgb image
Returns:
feat: Histogram of Gradient (HOG) feature
"""
# convert rgb to grayscale if needed
if im.ndim == 3:
image = rgb2gray(im)
else:
image = np.at_least_2d(im)
sx, sy = image.shape # image size
orientations = 9 # number of gradient bins
cx, cy = (8, 8) # pixels per cell
gx = np.zeros(image.shape)
gy = np.zeros(image.shape)
gx[:, :-1] = np.diff(image, n=1, axis=1) # compute gradient on x-direction
gy[:-1, :] = np.diff(image, n=1, axis=0) # compute gradient on y-direction
grad_mag = np.sqrt(gx ** 2 + gy ** 2) # gradient magnitude
grad_ori = np.arctan2(gy, (gx + 1e-15)) * (180 / np.pi) + 90 # gradient orientation
n_cellsx = int(np.floor(sx / cx)) # number of cells in x
n_cellsy = int(np.floor(sy / cy)) # number of cells in y
# compute orientations integral images
orientation_histogram = np.zeros((n_cellsx, n_cellsy, orientations))
for i in range(orientations):
# create new integral image for this orientation
# isolate orientations in this range
temp_ori = np.where(grad_ori < 180 / orientations * (i + 1),
grad_ori, 0)
temp_ori = np.where(grad_ori >= 180 / orientations * i,
temp_ori, 0)
# select magnitudes for those orientations
cond2 = temp_ori > 0
temp_mag = np.where(cond2, grad_mag, 0)
orientation_histogram[:,:,i] = uniform_filter(temp_mag, size=(cx, cy))[int(cx/2)::cx, int(cy/2)::cy].T
return orientation_histogram.ravel()