-
Notifications
You must be signed in to change notification settings - Fork 0
/
sudoku_project4.py
192 lines (169 loc) · 6.14 KB
/
sudoku_project4.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
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
import numpy as np
import cv2
from tensorflow.keras.models import load_model
from sudoku_solver import solve
from copy import deepcopy
import time
def put_digits(sudoku, N, sudoku_length, sudoku_height, x_coord, y_coord):
# x = x_coord/(max(sudoku_length, x_coord)/9)
# y = y_coord/(max(sudoku_height, y_coord)/9)
# x = min(8, x)
# y = min(8, y)
# sudoku[int(x)][int(y)] = N
x = x_coord/(sudoku_length/9)
y = y_coord/(sudoku_height/9)
sudoku[int(x)][int(y)] = N
print("put_digits")
def print_sudoku(copy):
# print("print_sudoku")
if(copy):
for i in range(9):
print(copy[i])
else:
print("cannot be solved")
# //////////////////////////////////////////////
cap = cv2.VideoCapture(0)
out = 0
model = load_model("computer_mnist_model100.h5") #("computer_mnist_model.h5") #("mnist_model.h5")
# //////////////////////////////////////////////
while(1):
# print("start")
sudoku = [[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0],
[0,0,0,0,0,0,0,0,0]]
ret, im = cap.read()
# print(im.shape)
imgray = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY) # BGR to grayscale
thresh = cv2.adaptiveThreshold(imgray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\
cv2.THRESH_BINARY_INV, 35, 12)
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE) # cv2.CHAIN_APPROX_NONE)
rects = [cv2.boundingRect(ctr) for ctr in contours]
if(not len(rects)):
continue
max_area_rect = 0
max_area_ctr = 0
max_area = 0
for ctr in contours:
rect = cv2.boundingRect(ctr)
if(rect[2]*rect[3] > max_area):
max_area = rect[2]*rect[3]
max_area_rect = rect
max_area_ctr = ctr
# if(abs(max_area_rect[2] - max_area_rect[3]) > 50
# or max_area_rect[3] < 280 or max_area_rect[2] < 280):
# continue
if(str(max_area_ctr) == '0'):
continue
epsilon = 0.01 * cv2.arcLength(max_area_ctr, True)
approx = cv2.approxPolyDP(max_area_ctr, epsilon, True)
im = cv2.drawContours(im, [approx], -1, (0, 255, 0), 1)
if(approx.shape[0] != 4):
print("73")
continue
else:
print("76")
up_left = approx[1][0]
up_right = approx[0][0]
bottom_left = approx[2][0]
bottom_right = approx[3][0]
# ======================================================================
# up_left = approx[0][0]
# up_right = approx[3][0]
# bottom_left = approx[1][0]
# bottom_right = approx[2][0]
input_pts = np.float32([up_left, up_right, bottom_left, bottom_right])
height = max(abs(bottom_left[1] - up_left[1]), abs(up_right[1]-bottom_right[1]))
width = max(abs(bottom_left[0] - bottom_right[0]), abs(up_left[0]-up_right[0]))
output_pts = np.float32([[0, 0], [width, 0],[0, height],[width, height]])
# =========================================================================================
# Compute the perspective transform M
M = cv2.getPerspectiveTransform(input_pts, output_pts)
# Apply the perspective transformation to the image
out = cv2.warpPerspective(im, M, (width, height), flags = cv2.INTER_LINEAR)
# print(up_left, up_right, bottom_left, bottom_right)
# print(width, height)
if(height > 220 and width > 220):# or abs(height - width) > 50):
thresh_out = cv2.warpPerspective(thresh, M, (width, height), flags = cv2.INTER_LINEAR)
edges = cv2.Canny(thresh_out, 1, 1, apertureSize = 3)
lines = cv2.HoughLines(edges,1,np.pi/180,100)
if(str(lines) == 'None'):
continue
for i in range(len(lines)):
for rho,theta in lines[i]:
a = np.cos(theta)
b = np.sin(theta)
x0 = a*rho
y0 = b*rho
x1 = int(x0 + (1000)*(-b)) # shape[0]
y1 = int(y0 + (1000)*(a)) # shape[1]
x2 = int(x0 - (1000)*(-b))
y2 = int(y0 - (1000)*(a))
if(abs(x1 - x2) > 50 and abs(y1 - y2) > 50):
continue
# cv2.line(img1, (x1,y1),(x2,y2),(255,255,255),3) # 2 points ,color, thikness
cv2.line(thresh_out, (x1,y1),(x2,y2),(0,0,0),3)
im_th = thresh_out
ctrs, hier = cv2.findContours(im_th.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
rects = [cv2.boundingRect(ctr) for ctr in ctrs]
sudoku_area = im_th.shape[0]*im_th.shape[1]//81
digits_array = []
for rect in rects:
if(rect[2]*rect[3] < sudoku_area//15):
continue
cv2.rectangle(out, (rect[0], rect[1]), (rect[0] + rect[2], rect[1] + rect[3]), (0, 255, 0), 1)
from_x = max(0, rect[1] - rect[3]//4)
to_x = min(im_th.shape[0], rect[1] + rect[3] + rect[3]//4)
from_y = max(0, rect[0] - rect[2]//4)
to_y = min(im_th.shape[1], rect[0] + rect[2] + rect[2]//4)
nbr_img = im_th[from_x: to_x, from_y:to_y]/255
nbr_resized_disp = cv2.resize(nbr_img, (28, 28))
nbr_resized = nbr_resized_disp.reshape(-1, 28, 28, 1)
digits_array.append(nbr_resized)
pred_array = model.predict(np.array(digits_array).reshape(-1, 28, 28, 1)).argmax(axis = 1)
N = 0
for rect in rects:
if(rect[2]*rect[3] < sudoku_area//15):
continue
# cv2.putText(out, str(pred_array[N]), (rect[0], rect[1]),cv2.FONT_HERSHEY_DUPLEX, 1, (200, 0, 0), 1)
if(out.shape[1] < rect[1] or out.shape[0] < rect[0]):
continue
# put_digits(sudoku, pred_array[N], out.shape[1], out.shape[0], rect[1], rect[0])
N += 1
if(N < 17):
print("not enough")
continue
sudoku_copy = deepcopy(sudoku)
# print("before solve")
# print_sudoku(sudoku)
# t = time.time()
# b = solve(sudoku_copy, 0, 0, sudoku_copy[0][0], t)
# print("after solve")
# print_sudoku(b)
if(str(b) == 'False'):
continue
# for i in range(9):
# for j in range(9):
# if(not sudoku[i][j]):
# print("putting", b[i][j], i, j)
# cv2.putText(out, str(b[i][j]), (int((j+1/3)*out.shape[0]//9), int((i+2/3)*out.shape[1]//9)), cv2.FONT_HERSHEY_PLAIN, 2, (0, 0, 255), 1)
# cv2.imshow("out", out)
# cv2.imshow("thresh_out", thresh_out)
# cv2.imshow("thresh", thresh)
if(str(out) == "0"):
continue
# print(out.shape[0], out.shape[1], approx)
cv2.imshow("out", out)
cv2.imshow("title", im)
# cv2.imshow("thresh", thresh)
# print(out.shape)
k = cv2.waitKey(30) & 0xff
if k == 27: # 27 is ascii code for ESC
break
cap.release()
cv2.destroyAllWindows()