def select_lines(self):
image_gray = img_fun.image_gray(self.image)
image_bin = cv2.adaptiveThreshold(image_gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, 33, 5)
horizontalsize = 50
horizontalStructure = cv2.getStructuringElement(cv2.MORPH_RECT, (horizontalsize, 1))
image_bin = cv2.erode(image_bin, horizontalStructure, iterations=1)
image_bin = cv2.dilate(image_bin,horizontalStructure, iterations=1)
# image_bin = cv2.dilate(image_bin,np.ones((2,500)), iterations=1)
image_bin = cv2.dilate(image_bin,np.ones((2,200)), iterations=1)
img,reg,pos = self.select_horizontal_lines(self.image.copy(),image_bin)
lines,groups = npt.add_additional_lines(pos)
self.lines = lines
self.groups = groups
# cv2.imshow('lines',cv2.resize(image_bin, (1000, 750), interpolation=cv2.INTER_NEAREST))
# cv2.waitKey(0)
# cv2.imshow('aaaaaaaaaa',img)
img_fun.show_image('bin',image_bin)
img_fun.show_image('img' ,img)
def predict_ann(self):
filename = tkFileDialog.askopenfilename(initialdir='images/')
image = cv2.imread(filename)
if image is None:
return
image_bin = self.prepare_image(image)
groups = self.get_groups(image)
image_orig,selected_regions, positions,reg_details = imgFunctions.select_roi(image.copy(), image_bin, groups)
self.inputs = ann_fun.prepare_for_ann(selected_regions)
self.outputs = ann_fun.convert_output(self.alphabet)
self.sheet = np_fun.get_notes(image)
with open('saved_ann/trained_ann.pkl', 'rb') as input:
self.ann = pickle.load(input)
results = self.ann.predict(np.array(self.inputs, np.float32))
print 'results = ',results
results = ann_fun.display_result(results, self.alphabet)
results = self.replace_beams(results)
print results
self.chords = self.sheet.set_chords_duration(results[2:])
tkMessageBox.showinfo("Done", "Sheet loaded. Click on Play Sheet button.")
img_fun.show_image('Sheet',image)
def show_lines_and_notes(self):
image = self.image.copy();
for reg in self.note_regs.regions:
cv2.rectangle(image, (reg.x, reg.y), (reg.x+reg.w, reg.y+reg.h), (255, 0, 0), 2)
for reg in self.h_lines:
cv2.rectangle(image, (reg.x, reg.y), (reg.x+reg.w, reg.y), (0, 255, 0), 2)
img_fun.show_image('all',image)
def select_bar_lines(self):
image = self.image.copy()
image_bin = self.prepare_bar_lines()
img, contours, hierarchy = cv2.findContours(image_bin.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
regions = regs.Regions(self.groups)
for contour in contours:
x,y,w,h = cv2.boundingRect(contour)
if w < 35 and h>100:
r = reg.Region(x,y,w,h)
regions.add_region(r)
cv2.rectangle(image, (x, y), (x+w, y+h), (0, 255, 0), 2)
# cv2.imshow('bar lines',image)
img_fun.show_image('bars',image)
self.bar_lines_regs = regions
def prepare_image(self):
self.image_gray = img_fun.image_gray(self.image)
self.image_bin = cv2.adaptiveThreshold(self.image_gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, 33, 5)
image_bin = self.image_bin
image_bin = cv2.erode(image_bin,np.ones((3,3)),iterations=1)
image_bin = cv2.dilate(image_bin,np.ones((3,20)),iterations=1)
kernel = np.ones((12,12),np.uint8)
image_bin = cv2.morphologyEx(image_bin, cv2.MORPH_OPEN, kernel)
image_bin = cv2.erode(image_bin,np.ones((3,30)),iterations=1)
img_fun.show_image('bars',image_bin)
self.image_bin = image_bin
self.select_regions(self.image.copy(),image_bin)
def prepare_bar_lines(self):
image_gray = img_fun.image_gray(self.image.copy())
image_bin = cv2.adaptiveThreshold(image_gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, 33, 5)
image_bin = cv2.erode(image_bin,np.ones((3,3)),iterations=1)
image_bin = cv2.dilate(image_bin,np.ones((3,20)),iterations=1)
kernel = np.ones((12,12),np.uint8)
image_bin = cv2.morphologyEx(image_bin, cv2.MORPH_OPEN, kernel)
image_bin = cv2.dilate(image_bin,np.ones((20,1)),iterations=1)
# horizontalsize = 50
# horizontalStructure = cv2.getStructuringElement(cv2.MORPH_RECT, (1,horizontalsize))
# image_bin = cv2.erode(image_bin, horizontalStructure, iterations=1)
# image_bin = cv2.dilate(image_bin,horizontalStructure, iterations=1)
# cv2.imshow('bar lines',image_bin)
# cv2.waitKey(0)
img_fun.show_image('bars',image_bin)
return image_bin
def select_regions(self,image,image_bin):
img, contours, hierarchy = cv2.findContours(image_bin.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
regions = regs.Regions(self.groups)
for contour in contours:
x,y,w,h = cv2.boundingRect(contour)
# if w>6 and w<19 and h > 11 and h < 27:
#if w>6 and w<30and h > 11 and h < 40 and x > 250:
if x > 250:
print 'w,h',w,h
r = reg.Region(x,y,w,h)
regions.add_region(r)
cv2.rectangle(image, (x, y), (x+w, y+h), (0, 255, 0), 2)
# cv2.imshow('lines',image)
# cv2.waitKey(0)
img_fun.show_image('regions',image)
self.note_regs = regions
