-
Notifications
You must be signed in to change notification settings - Fork 0
/
spark_code.py
123 lines (95 loc) · 3.1 KB
/
spark_code.py
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
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
from pyspark import SparkContext
import cv2
import numpy as np
import math
import sys
from operator import add
import mapreduce as mp
sys.path.append('/usr/local/lib/python2.7/site-packages')
def blockshaped(arr, nrows, ncols):
h, w = arr.shape
return (arr.reshape(h//nrows, nrows, -1, ncols)
.swapaxes(1,2)
.reshape(-1, nrows, ncols))
def mapreduce_to_file(sc, mag, angle, noOfRowInBlock, noOfColInBlock, xBlockSize, yBlockSize):
# sc = SparkContext("local", "Simple App")
# convert array to numpy array
mag = np.asarray(mag)
angle = np.asarray(angle)
# divide array into blocks
mag = blockshaped(mag, 20, 20)
angle = blockshaped(angle, 20, 20)
# Parallelize numpy array of blocks
mag_rdd = sc.parallelize(mag)
angle_rdd = sc.parallelize(angle)
R = noOfRowInBlock
C = noOfColInBlock
# Execute mapper function (compute averages) on RDDs
mag_blocks = mag_rdd.map(mp.mapper)
angle_blocks = angle_rdd.map(mp.mapper)
opFlowOfBlocks = np.zeros((xBlockSize, yBlockSize, 2))
i = 0
j = 0
for x in mag_blocks.toLocalIterator():
opFlowOfBlocks[i][j][0] = x
j = j + 1
if j >= yBlockSize:
j = 0
i = i + 1
i = 0
j = 0
for x in angle_blocks.toLocalIterator():
opFlowOfBlocks[i][j][1] = x
j = j + 1
if j >= yBlockSize:
j = 0
i = i + 1
# Stop the Spark Context
# sc.stop()
return opFlowOfBlocks
def opflow_mapreduce(sc, mag, angle, noOfRowInBlock, noOfColInBlock, xBlockSize, yBlockSize):
# convert array to numpy array
mag = np.asarray(mag)
angle = np.asarray(angle)
a, b = mag.shape
mag_values = np.zeros((a*b,3))
ang_values = np.zeros((a*b,3))
i = 0
for index, value in np.ndenumerate(mag):
mag_values[i][0] = int(index[0])
mag_values[i][1] = int(index[1])
mag_values[i][2] = value
i = i + 1
i = 0
for index, value in np.ndenumerate(angle):
ang_values[i][0] = int(index[0])
ang_values[i][1] = int(index[1])
ang_values[i][2] = value
i = i + 1
rdd_mag_values = sc.parallelize(mag_values)
rdd_mag_values = rdd_mag_values.map(mp.generate_key_value)
rdd_mag_values = rdd_mag_values.reduceByKey(add)
rdd_ang_values = sc.parallelize(ang_values)
rdd_ang_values = rdd_ang_values.map(mp.generate_key_value)
rdd_ang_values = rdd_ang_values.reduceByKey(add)
opFlowOfBlocks = np.zeros((xBlockSize, yBlockSize, 2))
for item in rdd_mag_values.collect():
key = item[0]
val = item[1]
col = key%100000
row = key/100000
opFlowOfBlocks[row][col][0] = val
for item in rdd_ang_values.collect():
key = item[0]
val = item[1]
col = key%100000
row = key/100000
opFlowOfBlocks[row][col][1] = val
# thefile = open('opflowofblocks mag new approach.txt', 'w')
# for A in opFlowOfBlocks:
# for B in A:
# thefile.write("%s\n" % B[0])
# thefile.close()
# sc.stop()
# sys.exit()
return opFlowOfBlocks