From 2de000f8168e90b1fb23f7baa967c4e8b1b84063 Mon Sep 17 00:00:00 2001
From: Aleksey Lobanov <alekseylobanov1@gmail.com>
Date: Wed, 29 Jun 2016 03:07:16 +0300
Subject: [PATCH] Added generation of train and test data

---
 settings.py        |  7 +++++
 train-generator.py | 77 ++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 84 insertions(+)
 create mode 100644 settings.py
 create mode 100644 train-generator.py

diff --git a/settings.py b/settings.py
new file mode 100644
index 0000000..05d55ef
--- /dev/null
+++ b/settings.py
@@ -0,0 +1,7 @@
+#!/usr/bin/python
+# -*- coding: utf-8 -*-
+
+FONTS_TO_PROCESS   = ['Arial', 'Cousine', 'Comfortaa']
+FONTS_SIZES = [12, 13, 14, 16, 20, 24, 36, 48]
+SYMBOLS_TO_PROCESS = 'QWERTYUIOPASDFGHJKLZXCVBNMqwertyuiopasdfghjklzxcvbnm1234567890.,():'
+NUMBER_OF_TEST_EXAMPLES_PER_FONT = 30
\ No newline at end of file
diff --git a/train-generator.py b/train-generator.py
new file mode 100644
index 0000000..1770ecc
--- /dev/null
+++ b/train-generator.py
@@ -0,0 +1,77 @@
+#!/usr/bin/python
+# -*- coding: utf-8 -*-
+
+import random
+import tempfile
+from collections import Counter
+import json
+
+import PIL
+from PIL import ImageDraw, ImageFont
+
+import settings
+
+
+def generatePhrase():
+    return ''.join([random.choice(settings.SYMBOLS_TO_PROCESS + ' ') for _ in range(random.randint(5,10))])
+
+
+def getImageWithText(text, font_name, font_size=16):
+    cur_font = PIL.ImageFont.truetype(font_name, 20)
+
+    # Временная картинка, чтобы получить размер текста
+    tmp_img = PIL.Image.new('RGBA', (100,100), (255,255,255,0))
+    pil_draw = PIL.ImageDraw.Draw(tmp_img)
+    text_size = pil_draw.textsize(text, cur_font)
+
+    img_size = (text_size[0] + 10, text_size[1] + 10)
+    img = PIL.Image.new('RGB', img_size, (255,255,255))
+    pil_draw = ImageDraw.Draw(img)
+    pil_draw.text((5,5), text, fill=(0,0,0), font=cur_font)
+    return img
+
+
+def main():
+    temp_dir = tempfile.TemporaryDirectory()
+    temp_path = temp_dir.name
+
+    error_count = Counter()  # Количество ошибок по каждому из шрифтовы
+
+    img_index = 1
+    img_info = {}  # Сюда запишем информацию по каждому из индексов
+
+    # Пишем по одному разу каждую букву каждого шрифта
+    for font_name in settings.FONTS_TO_PROCESS:
+        for symbol in settings.SYMBOLS_TO_PROCESS:
+            for font_size in settings.FONTS_SIZES:
+                try:
+                    img = getImageWithText(symbol, font_name)
+                    img.save('{}/{}.png'.format(temp_path,img_index),'PNG')
+                    img_info[img_index] = {'font':font_name, 'symbol':symbol, 'size':font_size}
+                except IOError:
+                    error_count[font_name] += 1
+                img_index += 1
+
+    # Создаём тестовые данные: сколько-то фраз на каждый шрифт
+    for font_name in settings.FONTS_TO_PROCESS:
+        for _ in range(settings.NUMBER_OF_TEST_EXAMPLES_PER_FONT):
+            try:
+                font_size = random.choice(settings.FONTS_TO_PROCESS)
+                phrase = generatePhrase()
+                img = getImageWithText(phrase, font_name)
+                img.save('{}/{}.png'.format(temp_path,img_index),'PNG')
+                img_info[img_index] = {'font':font_name, 'phrase':phrase, 'size':font_size}
+            except IOError:
+                error_count[font_name] += 1
+            img_index += 1
+
+    with open('{}/info.json'.format(temp_path),'w') as f:
+        json.dump(img_info, f)
+
+    print('Data stores in {}'.format(temp_path))
+    if len(error_count) != 0:
+        print('Errors: ')
+        print(error_count)
+
+if __name__ == "__main__":
+    main()