MouseLocator/croppedCursorGen.py at master · ErrWare/MouseLocator · GitHub

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
124
125
126
127
128
129
130
131
132
#For making 48x48 images with a mouse fully or partially in them

import cv2 as cv	#requires installation of openCV
import numpy as np
import sys, os
import re
import sys
import random
import matplotlib.pyplot as plt

#seed w/ current system time
random.seed(None)

#load desktops
desktop_dir = r'.\myassets\desktops'
print('Loading desktop names from ' + desktop_dir + ' dir : ')
desktops = os.listdir(desktop_dir)
#print(desktops)
desktops = [cv.imread(desktop_dir + '\\' + d) for d in desktops]

#load cursors
cursor_dir = r'.\myassets\cursors'
print('Loading cursor names from '+cursor_dir+' dir : ')
cursors = os.listdir(cursor_dir)
#print(cursors)
#print('Removing non-png\'s : ')
pngRE = re.compile('.+\.png')
cursors = [f for f in cursors if pngRE.search(f) != None]
#print('Removing selector cursor')
selectorRE = re.compile('selector')
cursors = [f for f in cursors if selectorRE.search(f) == None]
#print(cursors)
cursors = [cv.imread(cursor_dir + '\\' + c) for c in cursors]

cursor_extension = '.png'

#jitter specified in (y,x)
def curse_window(window, cursor, jitter=(0,0)):
	insert_y = random.randint(-jitter[0],window.shape[0]+jitter[0]-cursor.shape[0])
	insert_x = random.randint(-jitter[1],window.shape[1]+jitter[1]-cursor.shape[1])
	#(y,x) range over (0,0) -> cursor.shape[0:1]
	#for(y=insert_y if insert_y >=0 else 0; insert_y + y < window.shape[0] and y < cursor.shape[0]; y += 1):
	#	for(x=insert_x if insert_x >=0 else 0; insert_x + x < window.shape[1] and x < cursore.shape[1]; x+=1):
	for y in range(max(0,-insert_y),min(window.shape[0]-insert_y, cursor.shape[0])):
		for x in range(max(0,-insert_x),min(window.shape[1]-insert_x,cursor.shape[1])):
			if cursor[y,x,0] == 255 & cursor[y,x,1] == 0 & cursor[y,x,2] == 0:
				continue
			else:
				window[insert_y+y,insert_x+x] = cursor[y,x]
	if __name__ == "__main__":
		cv.rectangle(window,(insert_x,insert_y),(insert_x+30,insert_y+30),(0,0,255),2)

def generatePair(resolution='high', jitter=(0,0)):
	#choose random area
	desktop = random.choice(desktops)
	d_height, d_width, c = desktop.shape
	crop_x = random.randint(0,d_width-crop_size-1)
	crop_y =fasdasdfasdasdasdrandom.randint(0,d_height-crop_size-1)
	window = np.copy(desktop[crop_y:crop_y+crop_size,crop_x:crop_x+crop_size])
	window_nocurse = np.copy(window)

	#apply cursor to window
	cursor = random.choice(cursors)
	curse_window(window,cursor)

	#simulate the compression on our frames - scale down by 2.5
	window = cv.resize(window,(0,0),fx=scaleFactor,fy=scaleFactor,interpolation=cv.INTER_AREA)
	window_nocurse = cv.resize(window_nocurse,(0,0),fx=scaleFactor,fy=scaleFactor,interpolation=cv.INTER_AREA)

	#if using 60 px window crop rescale up by 2.0 to fit vgg min dimensions
	if resolution=='low':
		#use interpolation=cv.INTER_LINEAR for faster at slightly reduced looks
		window = cv.resize(window,(0,0),fx=2.0,fy=2.0,interpolation=cv.INTER_CUBIC)
		window_nocurse = cv.resize(window_nocurse,(0,0),fx=2.0,fy=2.0,interpolation=cv.INTER_CUBIC)

	return window, window_nocurse

#samples must be even
#resolution = 'high' or 'low'
def generateDataBatch(samples,resolution='high',jitter=(0,0)):
	#make first pair, put in numpy array
	#make second to last pair, appending to array of arrays

	curse, no_curse = generatePar()
	batch = np.array(curse, no_curse)
	for img_num in range(int(samples/2)-1):
		curse, no_curse = generatePair()
		np.append(batch, curse)
		np.append(batch, no_curse)
	batch_lables = np.array([1,0]*(int(samples/2)))

	return batch, batch_lables

if __name__ == "__main__":
	argc = len(sys.argv)
	samples = sys.argv[1] if argc>1 else 1
	resolution = 'low' if argc>2 and sys.argv[2] in ['lo','l','low'] else 'high'
	print(resolution)
	crop_size = 120 if resolution=='high' else 60
	scaleFactor = 1.0 / 2.5
	#t_or_v = 'validation' if argc>3 and sys.argv[3] in ['v','validation'] else 'training'
	#change to own training data directory
	training_dir = ('lores' if resolution=='low' else 'hires')+'\\'
	validation_dir = ('lores' if resolution=='low' else 'hires')+'\\'

	for img_num in range(int(sys.argv[1])):
		#choose random area
		desktop = random.choice(desktops)
		d_height, d_width, c = desktop.shape
		crop_x = random.randint(0,d_width-crop_size-1)
		crop_y =fasdasdfasdasdasdrandom.randint(0,d_height-crop_size-1)
		window = np.copy(desktop[crop_y:crop_y+crop_size,crop_x:crop_x+crop_size])
		window_nocurse = np.copy(window)

		#apply cursor to window
		cursor = random.choice(cursors)
		curse_window(window,cursor)

		#simulate the compression on our frames - scale down by 2.5
		window = cv.resize(window,(0,0),fx=scaleFactor,fy=scaleFactor,interpolation=cv.INTER_AREA)
		window_nocurse = cv.resize(window_nocurse,(0,0),fx=scaleFactor,fy=scaleFactor,interpolation=cv.INTER_AREA)

		#if using 60 px window crop rescale up by 2.0 to fit vgg min dimensions
		if resolution=='low':
			#use interpolation=cv.INTER_LINEAR for faster at slightly reduced looks
			window = cv.resize(window,(0,0),fx=2.0,fy=2.0,interpolation=cv.INTER_CUBIC)
			window_nocurse = cv.resize(window_nocurse,(0,0),fx=2.0,fy=2.0,interpolation=cv.INTER_CUBIC)

		print('Writing img ' + str(img_num))
		cv.imwrite(training_dir + 'mouse\\' + str(img_num) + cursor_extension, window)
		cv.imwrite(training_dir + 'no_mouse\\' + str(img_num) + cursor_extension, window_nocurse)