1# MIT License
2#
3# Copyright (c) 2017 OsciiArt
4#
5# Permission is hereby granted, free of charge, to any person obtaining a copy
6# of this software and associated documentation files (the "Software"), to deal
7# in the Software without restriction, including without limitation the rights
8# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9# copies of the Software, and to permit persons to whom the Software is
10# furnished to do so, subject to the following conditions:
11#
12# The above copyright notice and this permission notice shall be included in all
13# copies or substantial portions of the Software.
14#
15# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21# SOFTWARE.
22
23#!/usr/bin/env python
24# -*- coding: utf-8 -*-
25
26from keras.models import model_from_json
27import numpy as np
28import pandas as pd
29from PIL import Image
30import pickle
31import os
32
33
34# parameters
35model_path = "model/model.json"
36weight_path = "model/weight.hdf5"
37image_path = 'sample images/original images/21 original.png' # put the path of the image that you convert.
38new_width = 0 # adjust the width of the image. the original width is used if new_width = 0.
39input_shape = [64, 64, 1]
40
41
42def add_mergin(img, mergin):
43    if mergin!=0:
44        img_new = np.ones([img.shape[0] + 2 * mergin, img.shape[1] + 2 * mergin], dtype=np.uint8) * 255
45        img_new[mergin:-mergin, mergin:-mergin] = img
46    else:
47        img_new = img
48    return img_new
49
50
51def pickleload(path):
52    with open(path, mode='rb') as f:
53        data = pickle.load(f)
54    return data
55
56
57# load model
58json_string = open(model_path).read()
59model = model_from_json(json_string)
60model.load_weights(weight_path)
61print("model load done")
62
63char_list_path = "data/char_list.csv"
64char_list = pd.read_csv(char_list_path, encoding="cp932")
65print("len(char_list)", len(char_list))
66# print(char_list.head())
67char_list = char_list[char_list['frequency']>=10]
68char_list = char_list['char'].as_matrix()
69
70for k, v in enumerate(char_list):
71    if v==" ":
72        space = k
73        break
74print("class index of 1B space:", space)
75
76
77mergin = (input_shape[0] - 18) // 2
78img = Image.open(image_path)
79orig_width, orig_height = img.size
80if new_width==0: new_width = orig_width
81new_height = int(img.size[1] * new_width / img.size[0])
82img = img.resize((new_width, new_height), Image.LANCZOS)
83img = np.array(img)
84if len(img.shape) == 3:
85    img = img[:, :, 0]
86
87img_new = np.ones([img.shape[0]+2*mergin+18, img.shape[1]+2*mergin+18],
88                  dtype=np.uint8) * 255
89img_new[mergin:mergin+new_height, mergin:mergin+new_width] = img
90img = (img_new.astype(np.float32)) / 255
91
92char_dict_path = "data/char_dict.pkl"
93char_dict = pickleload(char_dict_path)
94
95print("len(char_dict)", len(char_dict))
96
97output_dir = "output/"
98if not os.path.isdir(output_dir):
99    os.makedirs(output_dir)
100
101for slide in range(18):
102    print("converting:", slide)
103    num_line = (img.shape[0] - input_shape[0]) // 18
104    img_width = img.shape[1]
105    new_line = np.ones([1, img_width])
106    img = np.concatenate([new_line, img], axis=0)
107    predicts = []
108    text = []
109    for h in range(num_line):
110        w = 0
111        penalty = 1
112        predict_line = []
113        text_line = ""
114        while w <= img_width - input_shape[1]:
115            input_img = img[h*18:h*18+ input_shape[0], w:w+input_shape[1]]
116            input_img = input_img.reshape([1,input_shape[0], input_shape[1], 1])
117            predict = model.predict(input_img)
118            if penalty: predict[0, space] = 0
119            predict = np.argmax(predict[0])
120            penalty = (predict==space)
121            char = char_list[predict]
122            predict_line.append(char)
123            char_width = char_dict[char].shape[1]
124            w += char_width
125            text_line += char
126        predicts.append(predict_line)
127        text.append(text_line+'\r\n')
128    # print(text)
129
130    img_aa = np.ones_like(img, dtype=np.uint8) * 0xFF
131
132    for h in range(num_line):
133        w = 0
134        for char in predicts[h]:
135            # print("w", w)
136            char_width = char_dict[char].shape[1]
137            char_img = 255 - char_dict[char].astype(np.uint8) * 255
138            img_aa[h*18:h*18+16, w:w+char_width] = char_img
139            w += char_width
140
141    img_aa = Image.fromarray(img_aa)
142    img_aa = img_aa.crop([0,slide,new_width, new_height+slide])
143    save_path = output_dir + os.path.basename(image_path)[:-4] + '_'\
144                + 'w' + str(new_width) \
145                + '_slide' + str(slide) + '.png'
146    img_aa.save(save_path)
147
148    f=open(save_path[:-4] + '.txt', 'w')
149    f.writelines(text)
150    f.close()
151print('http://example.com?a=')
152print('''http://example.com?a='b'&''')
153