mnist_cnn_classifier.py

import keras
import numpy as np
import cv2
from PIL import Image
from io import StringIO
import base64
from binascii import a2b_base64

#load pretrained model
model = keras.models.load_model("mnist_cnn_model.h5")

def data_uri_to_cv2_img(uri):
        encoded_data = uri.split(',')[1]
        # print("č")
        # img = base64.b64decode(uri)
        # nparr = np.fromstring(img, np.uint8)
        # img_np = cv2.imdecode(nparr, cv2.IMREAD_COLOR)  # cv2.IMREAD_COLOR in OpenCV 3.1
        data = encoded_data
        binary_data = a2b_base64(data)

        fd = open('imagelol.png', 'wb')
        fd.write(binary_data)
        fd.close()

        return cv2.imread('imagelol.png')

def add_border_to_image(im):
    WHITE = [255,255,255]

    bordersize = 100
    border = cv2.copyMakeBorder(im, top=bordersize, bottom=bordersize, left=bordersize, right=bordersize,
                                borderType=cv2.BORDER_CONSTANT, value=WHITE)
    return border



def cnn_predict(im):
    im = cv2.resize(im,(28,28))
    im=im/255
    predictedImage = np.resize(im,(1,28,28,1))
    prediction = model.predict_classes(predictedImage)
    return prediction


# Read the input image
im = cv2.imread("test.jpg")
#cv2.imwrite("base.png", im)
im = cv2.resize(im,(400,300))
im = add_border_to_image(im)

# Convert to grayscale and apply Gaussian filtering

# cv2.imwrite("base.png", im)

im_gray = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
im_gray = cv2.GaussianBlur(im_gray, (5, 5), 0)

# Threshold the image
ret, im_th = cv2.threshold(im_gray, 90, 255, cv2.THRESH_BINARY_INV)

# Find contours in the image
_,ctrs, hier = cv2.findContours(im_th.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cv2.imwrite("thresh.jpg", im_th)
# Get rectangles contains each contour
rects = [cv2.boundingRect(ctr) for ctr in ctrs]

# For each rectangular region, calculate HOG features and predict
# the digit using Linear SVM.
cnt=0
print(len(rects))
for rect in rects:
    cnt+=1
    #print(rect)
    # Draw the rectangles
    leng = int(rect[3]*1.4)
    pt1 = int(rect[1] + rect[3] // 2 - leng // 2)
    pt2 = int(rect[0] + rect[2] // 2 - leng // 2)
    roi = im_th[pt1:pt1 + leng, pt2:pt2 + leng]    # Make the rectangular region around the digit
    # Resize the image
   # roi = im_th[rect[1]:rect[1]+rect[3],rect[0]:rect[0]+rect[2]]
   # currim = im[y:y + h, x:x + w]
    print(roi.shape)
    x,y= roi.shape
    print(str(x) + " "  + str(y))
    print(int(x)>5 and int(y)>5)
    if(x>5 and y>5):
        cv2.rectangle(im, (rect[0], rect[1]), (rect[0] + rect[2], rect[1] + rect[3]), (0, 255, 0), 3)
        roi = cv2.resize(roi, (28, 28), interpolation=cv2.INTER_AREA)
        roi = cv2.dilate(roi, (3, 3))
        cv2.imwrite("im"+cnt.__str__()+".jpg",roi)
        nbr = cnn_predict(roi)
        cv2.putText(im, str(int(nbr[0])), (rect[0], rect[1]),cv2.FONT_HERSHEY_DUPLEX, 2, (0, 255, 255), 3)


cv2.imwrite("read_photo.jpg",im)
img = Image.open("read_photo.jpg")
img.show()