Combined2

# -*- coding: utf-8 -*-
"""
Created on Mon Jan 18 19:17:12 2021

@author: Sajid
"""

# -*- coding: utf-8 -*-
"""
Created on Mon Jan 18 18:33:48 2021

@author: Sajid

"""
from skimage.io import imread, imshow
from skimage.filters import gaussian, threshold_otsu
from skimage.feature import canny
from skimage.transform import probabilistic_hough_line, rotate

#testing
import numpy as np
import os
import cv2
import math
import matplotlib.pyplot as plt

import torch
from torch import nn
from torch import optim
import torch.nn.functional as F
from torchvision import datasets, transforms, models



from collections import OrderedDict
from PIL import Image

import pandas as pd
import seaborn as sns


# define the CNN architecture
class Net(nn.Module):
    ### TODO: choose an architecture, and complete the class

    def __init__(self):
        super(Net, self).__init__()

        # convolutional layer (sees 64x64x3 image tensor)
        self.conv1 = nn.Conv2d(3, 16, 3, padding=1)
        # convolutional layer (sees 32x32x16 tensor)
        self.conv2 = nn.Conv2d(16, 32, 3, padding=1)
        # convolutional layer (sees 16x16x32 tensor)
        self.conv3 = nn.Conv2d(32, 64, 3, padding=1)
        # convolutional layer (sees 8x8x64 tensor)
        self.conv4 = nn.Conv2d(64, 128, 3, padding=1)
        self.conv5 = nn.Conv2d(128, 256, 3, padding=1)

        # max pooling layer
        self.pool = nn.MaxPool2d(2, 2)
        # linear layer (256 * 2 * 2 -> 512)
        self.fc1 = nn.Linear(256 * 2 * 2 , 2048)
        # linear layer (512 -> 50)
        self.fc2 = nn.Linear(2048,512)
        # dropout layer (p=0.2)
        self.dropout = nn.Dropout(0.2)
        self.fc3 = nn.Linear(512,50)
        #self.softmax = nn.Softmax(dim=1)



    def forward(self, x):
        # add sequence of convolutional and max pooling layers
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = self.pool(F.relu(self.conv3(x)))
        x = self.pool(F.relu(self.conv4(x)))
        x = self.pool(F.relu(self.conv5(x)))
        # flatten image input
        x = x.view(-1, 256*2*2)

        # add dropout layer
        x = self.dropout(x)
        # add 1st hidden layer, with relu activation function
        x = F.relu(self.fc1(x))
        # add dropout layer
        x = self.dropout(x)
        # add 2nd hidden layer, with relu activation function
        x = self.fc2(x)
        x = self.dropout(x)
        x = self.fc3(x)
        return x
train_on_gpu = torch.cuda.is_available()
'''
if not train_on_gpu:
    print('CUDA is not available.  Training on CPU ...')
else:
    print('CUDA is available!  Training on GPU ...')
'''

classes=['অ','আ','ই', 'ঈ', 'উ','ঊ','ঋ','এ','ঐ', 'ও' ,   'ঔ','ক', 'খ',   'গ',    'ঘ',    'ঙ',    'চ',    'ছ',    'জ',    'ঝ',    'ঞ',    'ট',
    'ঠ',    'ড',    'ঢ',    'ণ',    'ত',    'থ',    'দ',    'ধ',    'ন',        'প',    'ফ',    'ব',    'ভ',    'ম',    'য',    'র',        'ল',                'শ' ,   'ষ',    'স' ,   'হ' ,
    'ড়','ঢ়','য়','ৎ','৹', ':', '৺']

model_scratch = Net()
model_scratch.load_state_dict(torch.load('model_scratch.pt' , map_location=torch.device('cpu')))


def process_image(image):
    ''' Scales, crops, and normalizes a PIL image for a PyTorch model

    '''

    img = Image.open(image)
    transformation = transforms.Compose([transforms.Resize([64,64]),
                                      #transforms.Grayscale(num_output_channels=1),
                                      transforms.ToTensor(),
                                      transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
                                      ])
    return transformation(img)

def predict(image_path, model):
    ''' Predict the class (or classes) of an image using a trained deep learning model.
    '''
    model.to('cpu')
    image = process_image(image_path)
    image = image.unsqueeze_(0)

    model.eval()
    with torch.no_grad():
        output = model.forward(image)

    #probabilities = torch.exp(output)

    #topk_probabilities, topk_labels = probabilities.topk(topk)
    #return([topk_probabilities, topk_labels]
    # convert output probabilities to predicted class
    _, preds_tensor = torch.max(output, 1)
    preds = np.squeeze(preds_tensor.numpy()) if not train_on_gpu else np.squeeze(preds_tensor.cpu().numpy())
    return(classes[preds])

def deskew(image):


    #threshold to get rid of extraneous noise
    thresh = threshold_otsu(image)
    normalize = image > thresh

    # gaussian blur
    blur = gaussian(normalize, 3)

    # canny edges in scikit-image
    edges = canny(blur)

    # hough lines
    hough_lines = probabilistic_hough_line(edges)

    # hough lines returns a list of points, in the form ((x1, y1), (x2, y2))
    # representing line segments. the first step is to calculate the slopes of
    # these lines from their paired point values
    slopes = [(y2 - y1)/(x2 - x1) if (x2-x1) else 0 for (x1,y1), (x2, y2) in hough_lines]

    # it just so happens that this slope is also y where y = tan(theta), the angle
    # in a circle by which the line is offset
    rad_angles = [np.arctan(x) for x in slopes]

    # and we change to degrees for the rotation
    deg_angles = [np.degrees(x) for x in rad_angles]

    # which of these degree values is most common?
    histo = np.histogram(deg_angles, bins=180)

    # correcting for 'sideways' alignments
    rotation_number = histo[1][np.argmax(histo[0])]

    if rotation_number > 45:
        rotation_number = -(90-rotation_number)
    elif rotation_number < -45:
        rotation_number = 90 - abs(rotation_number)

    return rotation_number



def deskew2(img,angle):

    #load in grayscale:


    #invert the colors of our image:
    cv2.imshow('input',img)
    cv2.bitwise_not(img, img)

    #compute the minimum bounding box:
    non_zero_pixels = cv2.findNonZero(img)
    center, wh, theta = cv2.minAreaRect(non_zero_pixels)

    root_mat = cv2.getRotationMatrix2D(center, angle, 1)
    rows, cols = img.shape
    rotated = cv2.warpAffine(img, root_mat, (cols, rows), flags=cv2.INTER_CUBIC)


    #Border removing:
    sizex = np.int0(wh[0])+10
    sizey = np.int0(wh[1])+10
    print(theta)
    if theta > -45 :
        temp = sizex
        sizex= sizey
        sizey= temp
    return cv2.getRectSubPix(rotated, (sizey,sizex), center)

#def breakword(img):
def detectlines(img):
    kernel = np.zeros((5,5),np.uint8)
    img = cv2.erode(img,kernel,iterations = 1)
    cv2.imshow('thinned',img)
    cv2.imwrite("noyse.jpg",img)
    thresh = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]

    # use morphology erode to blur horizontally
    ho,wo=img.shape
    ho=math.floor(ho/5)
    wo=math.floor(ho/10)

    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (ho,wo))
    morph = cv2.morphologyEx(thresh, cv2.MORPH_DILATE, kernel)

    # use morphology open to remove thin lines from dotted lines
    #kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 17))
    #morph = cv2.morphologyEx(morph, cv2.MORPH_OPEN, kernel)

    # find contours
    cntrs = cv2.findContours(morph, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    cntrs = cntrs[0] if len(cntrs) == 2 else cntrs[1]

    print('yo')
    print(len(cntrs))
    # Draw contours excluding the topmost box
    result = img.copy()
    ara = []
    lng=len(cntrs)
    for i in range(0,lng):
        c=cntrs[i]
        box = cv2.boundingRect(c)
        x,y,w,h = box
        cv2.rectangle(result, (x, y), (x+w, y+h), (0, 0, 255), 2)
        cur = img[y:y+h,x:x+w].copy()
        ara.append(cur)
        '''
        if len(cntrs)==1:
            ara.append(cur)
        else:
            ara1 = detectlines(cur)
            for i in ara1:
                ara.append(i)
        '''


    # write result to disk
    print(len(ara))
    sz=len(ara)


    cv2.imwrite("text_above_lines_threshold.png", thresh)
    cv2.imwrite("text_above_lines_morph.png", morph)
    cv2.imwrite("text_above_lines_lines.jpg", result)

    #cv2.imshow("GRAY", gray)
    cv2.imshow("THRESH", thresh)
    cv2.imshow("MORPH", morph)
    cv2.imshow("RESULT", result)
    return ara


img = cv2.imread(r"D:\Bangla OCR Dataset\Dataset\Dataset\1\1_3.jpg",cv2.IMREAD_GRAYSCALE)
cv2.imshow('input image',img)

print(img.shape)
img = cv2.adaptiveThreshold(img,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C,cv2.THRESH_BINARY,11,2)
cv2.imshow('binary image',img)

angel = deskew(img)
img = deskew2(img,angel)
cv2.bitwise_not(img,img)
cv2.imshow("Skew Corrected",img)

img = cv2.fastNlMeansDenoising(img, img, 50.0, 7, 21)
cv2.imshow('noiseless image1',img)
#img = cv2.morphologyEx(img, cv2.MORPH_OPEN, kernel)
#cv2.imshow('noiseless image2',img)

# threshold the grayscale image
ara=detectlines(img)
sz=len(ara)
print('yoyo')
print(sz)
sliced = []
for i in range(0,sz):
    cv2.imshow('Crop %d' % (i,), ara[sz-i-1])
    cv2.imwrite("Crp%d.png" % (i,), ara[sz-i-1])
    img2=ara[sz-i-1]
    thresh2 = cv2.threshold(img2, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
    print(img2.shape)
    he,we=img2.shape
    he =math.ceil(he/5.0)
    we = math.floor(we/10.0)
    print("ksdljjjjj")
    print(ara[sz-i-1].shape)
    # use morphology erode to blur horizontally
    kernel2 = cv2.getStructuringElement(cv2.MORPH_RECT, (he,we))
    morph2 = cv2.morphologyEx(thresh2, cv2.MORPH_DILATE, kernel2)

    # use morphology open to remove thin lines from dotted lines
    #kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 17))
    #morph = cv2.morphologyEx(morph, cv2.MORPH_OPEN, kernel)

    # find contours
    cntrs2 = cv2.findContours(morph2, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    cntrs2 = cntrs2[0] if len(cntrs2) == 2 else cntrs2[1]


    # Draw contours excluding the topmost box

    result2 = img2.copy()
    temp=[]
    for c2 in cntrs2:
        box2 = cv2.boundingRect(c2)
        x,y,w,h = box2
        cv2.rectangle(result2, (x, y), (x+w, y+h), (0, 0, 255), 2)
        temp.append(img2[y:y+h,x:x+w].copy())
    cv2.imwrite("temporary%d.png" % (i), result2)
    sliced.append(temp)
    sz2=len(temp)
    print(sz2)
for i in range(0,sz):
    sz2=len(sliced[sz-i-1])
    for j in range(0,sz2):
        sliced[sz-i-1][sz2-j-1]=cv2.inRange(sliced[sz-i-1][sz2-j-1],221,255)
        sliced[sz-i-1][sz2-j-1]=cv2.merge([sliced[sz-i-1][sz2-j-1],sliced[sz-i-1][sz2-j-1],sliced[sz-i-1][sz2-j-1]])
        cv2.imshow('sliced6%d%d' % (sz-i-1,j), sliced[sz-i-1][sz2-j-1])
        cv2.imwrite("sli9%d%d.png" % (sz-i-1,j), sliced[sz-i-1][sz2-j-1])
        print('yeah')
        print(sliced[sz-i-1][sz2-j-1].shape)
        #breakword(sliced[sz-i-1][sz2-j-1])

#print(predict('C:/Users/Sajid/bcc000010.bmp',model_scratch))
cv2.waitKey(0)