def analyze_img(filename):
"""
It will pass the image through the whole pipeline of pre-processing, segmentation, and classification.
Args:
filename (str): Name of the image to be processed
Returns:
Translated text in a list
"""
preprocessor = Preprocessor()
mynet = MyNetwork()
binarized_img = preprocessor.binarize(filename)
binarized_img = preprocessor.removeExtras(binarized_img)
binarized_img = preprocessor.enlargeLetters(binarized_img)
binarized_img = preprocessor.despeckle(binarized_img)
# # To visualize binarized image
# cv.namedWindow("test", cv.WINDOW_NORMAL)
# cv.imshow("test", binarized_img)
# cv.waitKey(1000)
_, image_lines = segmenter.dilateToSegments(binarized_img,
filename,
save_img_with_bb=False)
translated_text = []
for l, line in enumerate(image_lines):
translated_line = []
for i, component in enumerate(line):
print("Starting element {}/{} of line {}/{}".format(
i + 1, len(line), l + 1, len(image_lines)))
letter_images = segmenter.cluster_letters(component)
for letter in letter_images:
# # Visualize letter that will go to the NN
# cv.namedWindow("test", cv.WINDOW_NORMAL)
# cv.imshow("test", letter)
# cv.waitKey(1000)
probabilities, predicted_letter = mynet.predict_image(letter)
print("Predicted letter: {}".format(predicted_letter))
translated_line.append(predicted_letter)
translated_text.append(translated_line)
print("Translated line is: {}".format(translated_line))
return translated_text
from preprocessor import Preprocessor
import segmenter
from recognizer import Recognizer
import cv2 as cv
import os
preprocessor = Preprocessor()
# preprocessor.dilateToSegments("./binarized-images/P583-Fg006-R-C01-R01-fused.jpg")
for filename in os.listdir("./image-data"):
if filename.endswith("fused.jpg"):
image_binarized = preprocessor.binarize("./image-data/" + filename)
# image_binarized = preprocessor.cropImage(image_binarized)
image_binarized = preprocessor.despeckle(image_binarized)
image_binarized = preprocessor.removeExtras(image_binarized)
image_binarized = preprocessor.enlargeLetters(image_binarized)
cv.imwrite("./binarized-images/" + filename,
image_binarized) # Save test image for testing purposes
segmenter.print_histogram("./binarized-images/" + filename)
linesOfBoundingBoxes = segmenter.dilateToSegments(
image_binarized, filename)
else:
continue
'''
#image_binarized = preprocessor.binarize("./image-data/P344-Fg001-R-C01-R01-fused.jpg")
image_binarized = preprocessor.binarize("./image-data/P632-Fg002-R-C01-R01-fused.jpg")
image_binarized = preprocessor.cropImage(image_binarized)
