# -*- coding:utf-8 -*-

import keras
from keras.applications.vgg19 import VGG19
from keras.applications.densenet import decode_predictions
from PIL import Image
import numpy as np
import os.path
from DenseNet.DenseNet_ccc import densenet
from keras.models import Model
from keras.layers import Conv2D, Dense, Input, add, Activation, AveragePooling2D, GlobalAveragePooling2D, Lambda, \
    concatenate
from keras import optimizers, regularizers
from matplotlib import pyplot as plt
import cv2
import pickle
import copy


def count_paint(path):   # 用于提示汉字的剪切
    img = cv2.imread(path)  # 读取图片，装换为可运算的数组
    img = img[344:384, 0:344]
    GrayImage = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)  # 将BGR图转为灰度图
    ret, thresh1 = cv2.threshold(GrayImage, 160, 255, cv2.THRESH_BINARY)  # 将图片进行二值化（130,255）之间的点均变为255（背景）

    (h, w) = thresh1.shape  # 返回高和宽
    a = [0 for z in range(0, w)]
    # # 记录每一列的波峰
    for j in range(0, w):  # 遍历一列
        for i in range(0, h):  # 遍历一行
            if thresh1[i, j] == 0:  # 如果改点为黑点
                a[j] += 1  # 该列的计数器加一计数

    b = []
    for j in range(0, w):
        if a[j] == 0:
            b.append(j)

    c = []
    for i in range(len(b)-1):
        if b[i+1] - b[i] >= 3:
            c.append(b[i])
            c.append(b[i+1])

    d = []
    for i in range(1, len(c) - 1, 2):
        if c[i + 1] - c[i] < 2:
            d.append(i)
    if len(d) != 0:
        d.reverse()
        for i in range(len(d)):
            if c[d[i] + 2] - c[d[i] - 1] <= 40:
                c.pop(d[i] + 1)
                c.pop(d[i])

    e = []
    for i in range(0, len(c), 2):
        if c[i + 1] - c[i] < 10:
            left = c[i + 1] - c[i - 2] if i - 2 >= 0 else 40
            right = c[i + 3] - c[i] if i + 3 <= len(c) else 40
            if left < right:
                e.append(i - 1)
                e.append(i)
            else:
                e.append(i + 1)
                e.append(i + 2)
    if len(e) != 0:
        e.reverse()
        for i in range(len(e)):
            c.pop(e[i])

    imgs = []
    for i in range(0,len(c),2):
        sss = thresh1[0:40, c[i]:c[i+1]]
        size = sss.shape
        if size[1] < 40:
            padding = 40-size[1]
            left = padding//2
            sss = cv2.copyMakeBorder(sss, 0, 0, left, padding-left, cv2.BORDER_CONSTANT, value=[255,255,255])
            imgs.append(sss)
        else:
            print(size,'++++++++')
            plt.imshow(sss, cmap=plt.gray())
            plt.show()
            imgs.append(sss)
    return imgs, thresh1

def get_label_dict():
    f=open('./chinese_labels','rb')
    label_dict = pickle.load(f)
    f.close()
    return label_dict

img_rows, img_cols = 40, 40
img_channels = 1
num_classes = 3755
img_input = Input(shape=(img_rows, img_cols, img_channels))
output = densenet(img_input, num_classes)
model = Model(img_input, output)
model.load_weights('model-ep008-loss0.615-val_loss0.633.h5')
sgd = optimizers.SGD(lr=.1, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])
label_dict = get_label_dict()
path = '/Users/haodexian/PycharmProjects/scrapy-async-selenium-develop/reviews/img/'
img_generator = (os.path.join(path, file) for file in os.listdir(path))

while True:
    text = input('input "n" ( input Q to exit ):  ')
    if text == 'Q' :
        break
    if text == 'n' :
        img_path = next(img_generator)
        print(img_path)
    if not os.path.exists(img_path):
        print("file not exist!")
        continue
    try:
        images, ori = count_paint(img_path)
        plt.imshow(ori[:,0:150], cmap=plt.gray())
        plt.show()
        for i in range(len(images)):
            img = cv2.resize(images[i], (40, 40))
            img = (img.reshape(1, 40, 40, 1)).astype('int32')/255  #例子的个数，图片的 通道数，图片的长与宽
            results = model.predict(img)[0]
            # predict = np.argmax(results, axis=1)
            # print(label_dict[predict[0]])
            top_indices = results.argsort()[-20:][::-1]
            result = []
            for i in top_indices:
                if results[i] * 100 >= 1:
                    result.append((label_dict[i],int(results[i] * 100)))
            print(result)
    except Exception as e:
        raise e