def generateData(vis=False):
chars = string.ascii_uppercase + string.digits
# chars = string.digits
font = np.arange(0, 6, 1) #randint(0, 5)
# size = np.arange(2.5, 4, 0.5) #uniform(2.5, 3.5)
line_size = np.arange(1, 4, 1) #randint(1, 3)
blur = np.arange(0, 2, 1) #randint(0, 1)
kw = np.arange(1, 9, 2) #randint(3, 7)
kh = np.arange(1, 9, 2) #randint(3, 7)
generatedImgs = []
rows = []
for c in chars:
print("Generating - ", c)
row = []
for f in font:
for l in line_size:
# for b in blur:
for i in kw:
for j in kh:
img, _ = genSymbolImg(c, f, l, 1, i, j)
# _, croped = segmentSymbolsByContours(img)
resized = cv2.resize(img, (30, 40))
row.append(resized)
# cv2.imshow("W", resized)
# k = cv2.waitKey(0)
# if k == 27:
# exit()
generatedImgs.append(row)
rows.append(tl.concat_hor(row))
# cv2.imshow("ROW", tl.concat_hor(row))
# k = cv2.waitKey(0)
# if k == 27:
# break
print("Generation done")
generatedVisImg = tl.concat_ver(rows)
return generatedImgs, generatedVisImg
import cv2
import numpy as np
import tools as tl
img = cv2.imread("sudoku.png", 0)
ret, th = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY)
ret, th_inv = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY_INV)
th_adap11 = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2)
th_adap5 = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 11, 2)
ret_o, th_o = cv2.threshold(img, 0, 255,
cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
print(ret_o)
print(cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
row1 = tl.concat_hor((img, th, th_o))
row2 = tl.concat_hor((img, th_adap11, th_adap5))
final_img = tl.concat_ver((row1, row2))
cv2.imshow("Sudoku", final_img)
cv2.waitKey(0)
textRegions = []
for box in reversed(boxRegions):
if box[3] > meanHeight:
x, y, w, h = box
textRegions.append((x, y, x + w, y + h))
cv2.rectangle(vis, (x, y), (x + w, y + h), (0, 255, 0), 1)
textImg = []
for i in textRegions:
x1, y1, x2, y2 = i
textImg.append(img[y1:y2, x1:x2])
allTexts = tl.concat_ver(textImg)
row1 = tl.concat_hor((img, img_inv, filtered))
row2 = tl.concat_hor((th_o, closing, vis, allTexts))
# final = tl.concat_ver((row2))
cv2.imshow('myimg', cv2.resize(row2, (0, 0), fx=1, fy=1))
cv2.waitKey(0)
# for root, dirs, files in os.walk("../Data/"):
# for file in files:
# imagepath = os.path.join(root, file)
# if file.endswith('.jpg'):
# img = cv2.imread(imagepath)
# '''
# ToDo: find text areas
basicHeuristicAnalysisRanges = temp[temp[:, 4].argsort()]
basicHeuristicAnalysisRanges = basicHeuristicAnalysisRanges.astype(int)
'''''' '''''' '''''' '''''' '''
Deep Heuristic Analysis
''' '''''' '''''' '''''' ''''''
_verprojsBHA = np.zeros((1, img.shape[1]), dtype=np.uint8)
for r in basicHeuristicAnalysisRanges:
_band = gray[r[0]:r[1], r[2]:r[3]]
if _band.shape[0] > 0 and _band.shape[1] > 0:
bmin = np.min(_band)
bmax = np.max(_band)
bmid = int((bmax - bmin) / 2)
hist, bins = np.histogram(_band.ravel(), 256, [0, 256])
b2 = np.sum(hist[bmid:bmax]) - np.sum(hist[bmin:bmid])
# Debug Heuristic analysis
_img = img[r[0]:r[1], r[2]:r[3]]
_color = (0, 255, 0) if b2 > 0 else (0, 0, 255)
_img = cv2.copyMakeBorder(_img,
2,
2,
2,
2,
cv2.BORDER_CONSTANT,
value=_color)
_verprojsBHA = tl.concat_ver2((_verprojsBHA, _img))
showImg = tl.concat_hor((img, _verprojsBHA))
cv2.imwrite("debug_imgs/HBA/" + str(framenum) + ".jpg", showImg)
framenum += 1
verP = np.sum(gray, axis=1) / 255
verPVis = tl.getDrawProjectionVer(vis, verP)
gradX = cv2.Sobel(gray, ddepth=cv2.CV_64F, dx=1, dy=0, ksize=3)
gradX = np.abs(gradX)
(minVal, maxVal) = (np.min(gradX), np.max(gradX))
if maxVal - minVal > 0:
gradX = (255 * (
(gradX - minVal) / float(maxVal - minVal))).astype("uint8")
else:
gradX = np.zeros(gray.shape, dtype="uint8")
verPX = np.sum(gradX, axis=1) / 255
verPXVis = tl.getDrawProjectionVer(vis, verPX)
row1 = tl.concat_hor((gray, verPVis))
row2 = tl.concat_hor((gradX, verPXVis))
bigImg = tl.concat_ver((row1, row2))
scale = 0.8
cv2.imshow("Window2", cv2.resize(bigImg, (0, 0),
fx=scale,
fy=scale))
cv2.moveWindow("Window2", 0, 0)
k = cv2.waitKey(0)
if k == 27:
exit()
elif k == ord('q'):
exit()
import cv2
import numpy as np
import tools as tl
img = cv2.imread("sudoku.png", 0) # to be black-white
ret, th = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY)
ret, th_inv = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY_INV)
th_adap11 = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2)
th_adap5 = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
cv2.THRESH_BINARY, 5, 2)
#adaptive treshold divivedv the picture and нахожит порог для каждого
#поэтому он ничего не возвращает
ret_0, th_0 = cv2.threshold(img, 0, 255,
cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
print(ret_0)
print(cv2.THRESH_BINARY + cv2.THRESH_OTSU)
row1 = tl.concat_hor((img, th, th_0))
row2 = tl.concat_hor((img, th_adap5, th_adap11))
final_img = tl.concat_ver((row1, row2))
cv2.imshow("Sodoku", final_img)
cv2.waitKey(0)
# draw peaks
for peak in peaks:
cv2.circle(drawedverpConvolved, peak, 2, (0, 0, 255), -1)
# draw bands
bandsImg = np.zeros(vis.shape, dtype=np.uint8)
for band in bandP1ranges:
yt, yb = band
bandsImg[yt:yb] = [0, 255, 0]
vis = cv2.addWeighted(vis, 0.6, bandsImg, 0.4, 0)
'''
Find peaks Row Clipping Phase 2
'''
# TODO crop words using vertical projections
'''
Final drawing
'''
scale1 = 0.8
bigImg1 = tl.concat_hor(
(vis, grad1, drawedverp, sliceLine, drawedverpConvolved))
cv2.imshow("Window1", cv2.resize(bigImg1, (0, 0), fx=scale1, fy=scale1))
k = cv2.waitKey(0)
if k == 27:
exit()
elif k == ord('q'):
exit()
if r[6]:
_color = (0, 255, 0)
if globalFlagDetected == False:
if r[5] == maxDensity:
_color = (0, 255, 0)
elif r[5] == secondMaxDensity:
_color = (212, 255, 127)
elif r[5] == thirdMaxDensity:
_color = (255, 255, 0)
if r[4] < 260:
_color = (0, 191, 255)
if len(basicHeuristicAnalysisRanges) == 1:
_color = (0, 255, 0)
_img = tl.concat_hor((_img, r[7], r[8]))
_img = cv2.copyMakeBorder(_img,
2,
2,
2,
2,
cv2.BORDER_CONSTANT,
value=_color)
_verprojsBHA = tl.concat_ver2((_verprojsBHA, _img))
showImg = tl.concat_hor((img, _verprojsBHA))
cv2.imwrite("debug_imgs/HBA/" + str(framenum) + ".jpg", showImg)
framenum += 1
if box[3] > meanHeight:
x,y,w,h = box
textRegions.append((x,y, x+w,y+h))
cv2.rectangle(vis,(x,y),(x+w,y+h),(0,0,255),1)
'''
Extract all text regions
'''
regionImgs = []
for textRegion in textRegions:
x1,y1,x2,y2 = textRegion
regionImgs.append(img[y1:y2,x1:x2])
scale = 0.8
row1 = tl.concat_hor((vis, gradX, blur))
row2 = tl.concat_hor((th, opening))
allTexts = tl.concat_ver(regionImgs)
bigImg = tl.concat_ver((row1, row2))
cv2.imshow("Window1", cv2.resize(bigImg, (0,0), fx = scale, fy = scale))
cv2.imshow("Window2", cv2.resize(allTexts, (0,0), fx = 1.4, fy = 1.4))
k = cv2.waitKey(0)
if k == 27:
exit()
elif k == ord('q'):
exit()
