paper---cls实现过程记录

一、把video切分为tile

1. 把video切分成以秒为单位的segment

https://docs.opencv.org/3.0-beta/doc/py_tutorials/py_gui/py_video_display/py_video_display.html

2.把segment切分成rows*colums个tiles

二、 把user的fixaction进行聚类

三、使用svc进行预测

---

错误整理

1）

图1. 生成video时的编码

解决方法：

Python / OpenCV 2.4.9 doesn't support cv2.VideoWriter_fourcc, which is version 3.x. If you're using 2.4.x:

replace fourcc = cv2.VideoWriter_fourcc(*'XVID')

with fourcc = cv2.cv.CV_FOURCC(*'XVID')

2）

错误：

                        for r in range(0,rows):
                            for c in range(0,colums-1):
                                img = frame[r*image_height:r*image_height+image_height,c*image_width:c*image_width+image_width]                         
                                tile_frame.append(img)
                            # the last colum needs to be processed in different ways
                            img_last_col = frame[r*image_height:r*image_height+image_height,(c)*image_width:]
                            tile_frame.append(img_last_col)

执行结果：

正确：

                        for r in range(0,rows):
                            for c in range(0,colums-1):
                                img = frame[r*image_height:r*image_height+image_height,c*image_width:c*image_width+image_width]                         
                                tile_frame.append(img)
                            # the last colum needs to be processed in different ways
                            img_last_col = frame[r*image_height:r*image_height+image_height,(c+1)*image_width:]
                            tile_frame.append(img_last_col)

执行结果：分析c在执行最内层的循环 for c in range(0,colums-1) 后，c的值为colums-2，不是colums-1；然而，img_last_col对应的colum应该是colums-1.

3）导入特定path下的module

import sys
sys.path.append("F:/project/python")
from utils import WRFile # utils是文件名

4）python中extend和append的区别

a = []
item = ["time"]
item.extend([1,2,4,5])
a.append(item)
print(a)

5）常见的字符串操作

6）string批量转float

import numpy as np
a = ["wj","1.3","2.1"]
b = np.array(a[1:],dtype = "float64")
print(b) #[1.3 2.1]

7）list.append的返回值情况

a = ["wj","1.3","2.1"]
c = []
d = c.append(a) # 返回结果为None
print(c,d)  #[['wj', '1.3', '2.1']]   None

8）numpy创建多维度数组

import numpy as np
a = np.array([[[1,2,3],[4,5,6]],[[7,8,9],[10,11,12]],[[13,14,15],[16,17,18]]])
print(a)

[[[ 1  2  3]
  [ 4  5  6]]

 [[ 7  8  9]
  [ 0  0  0]]

 [[13 14 15]
  [16 17 18]]]

修改特定元素的值

a[1,1,:] = np.zeros(3)
print(a)

[[[ 1  2  3]
  [ 4  5  6]]

 [[ 7  8  9]
  [ 0  0  0]]

 [[13 14 15]
  [16 17 18]]]

水平增加一维度

垂直增加一维度

import numpy as np
b = np.arange(1,11,1).reshape((5,2))
c = np.zeros((1,2))
d = np.ones((5,1))
f = np.vstack((b,c)) # 在b的纵向增加元素
e = np.hstack((b,d))#在b的横向增加元素
print("=========vstack==========")
print(f)
print("=========hstack==========")
print(e)

输出结果

=========vstack==========
[[ 1.  2.]
 [ 3.  4.]
 [ 5.  6.]
 [ 7.  8.]
 [ 9. 10.]
 [ 0.  0.]]
=========hstack==========
[[ 1.  2.  1.]
 [ 3.  4.  1.]
 [ 5.  6.  1.]
 [ 7.  8.  1.]
 [ 9. 10.  1.]]

统计numpy中某个元素出现的次数

labels = clustering.labels_ #待分析的numpy数组

mask0 = (labels==0) #统计label内个元素是否等于0，如果等于，mask0的相应位置去True，反之为false
mask1 = (labels == 1)
mask_1 = (labels ==-1)
 count_1 = labels[mask_1].size #获得mask_1取True时，labels的元素
 count0 = labels[mask0].size
count1 = labels[mask1].size

生成[a,b]内n个不重复的随机数

import numpy as np

import random as ram

a = np.arange(0,48,1)

b = ram.sample(range(0,48),43)

根据随机生成的索引读取数据

import numpy as np
import random as ram
data = np.arange(1,16,1).reshape(5,3)
whole = range(0,5)
samples = ram.sample(whole,3)
test = list(set(whole)-set(samples))#把set转化为list
print("samples",samples)
print("test ",test)

print("data")
print(data)
print("sampele data")
print(data[samples])
print("test data")
print(data[test])

9）分析向量的欧氏距离，确定参数eps

            dist =  DistanceMetric.get_metric('euclidean')
            result = dist.pairwise(X)

10）绘图

    def plotScatter(self,x,y,z,slot):
        plt.scatter(x=x,y=y,c=z)
        plt.xlabel("x",fontsize=20)
        plt.ylabel("y",fontsize=20)
        plt.legend("z",fontsize=20)
        plt.show()
        plt.savefig("F:\\project\\dataset\\vr\\Formated_Data\\Experiment_1\\image/xyz"+str(slot)+".png")