首页 探索 帮助
登录
andrew
/
xword
关注 1
点赞 0
派生 0

4 次代码提交 82ef7f2999 ... a62558add7

作者 SHA1 备注 提交日期
  Andrew Klopper a62558add7 Add HTML output. 6 年之前
  Andrew Klopper 6fdfd76066 Split extract_crossword into extract_crossword_grid and grid_to_png 6 年之前
  Andrew Klopper 3de3b2f972 Added ability to filter out non-greys before trying to extract the 6 年之前
  Andrew Klopper 8ab61aeb6d Added 'extract' command. 6 年之前
共有 14 个文件被更改，包括 293 次插入 和 81 次删除
分列视图 显示文件统计
  1. 43 0
      home/jinja2/home/grid.html
  2. 16 0
      home/jinja2/home/index.html
  3. 33 0
      home/jinja2/home/output.html
  4. 0 0
      home/management/__init__.py
  5. 0 0
      home/management/commands/__init__.py
  6. 104 0
      home/management/commands/extract.py
  7. 0 16
      home/templates/home/index.html
  8. 0 33
      home/templates/home/output.html
  9. 9 3
      home/views.py
  10. 67 27
      home/xword.py
  11. 2 0
      requirements.in
  12. 4 1
      requirements.txt
  13. 12 0
      xword/jinja2.py
  14. 3 1
      xword/settings.py

+ 43 - 0
home/jinja2/home/grid.html
查看文件 

@@ -0,0 +1,43 @@
1 
+<!DOCTYPE html>
2 
+<html>
3 
+    <head>
4 
+        <meta charset="UTF-8">
5 
+        <title>Crossword</title>
6 
+        <style>
7 
+            .crossword {
8 
+                border-right: 1px solid;
9 
+                border-bottom: 1px solid;
10 
+            }
11 
+            .crossword td {
12 
+                border-left: 1px solid;
13 
+                border-top: 1px solid;
14 
+            }
15 
+            .crossword td.filled {
16 
+                background: black;
17 
+            }
18 
+        </style>
19 
+    </head>
20 
+    <body>
21 
+        <table border=0 cellspacing=0 cellpadding=0 class="crossword">
22 
+            {%- set tab_index = namespace(value=1) %}
23 
+            {%- set blank = '&emsp;'|safe %}
24 
+            {%- for row in range(num_rows) %}
25 
+                <tr>
26 
+                    {%- for col in range(num_cols) %}
27 
+                        {%- if grid[row][col].filled %}
28 
+                            <td class="filled">&emsp;</td>
29 
+                        {%- else %}
30 
+                            <td class="empty" tabindex="{{ tab_index.value }}">
31 
+                                {{ grid[row][col].number|default(blank) }}
32 
+                            </td>
33 
+                            {%- set tab_index.value = tab_index.value + 1 %}
34 
+                        {%- endif %}
35 
+                    {%- endfor %}
36 
+                </tr>
37 
+            {%- endfor %}
38 
+        </table>
39 
+    </body>
40 
+</html>
41 
+<!--
42 
+vim:ts=4:sw=4:expandtab
43 
+-->

+ 16 - 0
home/jinja2/home/index.html
查看文件 

@@ -0,0 +1,16 @@
1 
+<!DOCTYPE html>
2 
+<html>
3 
+    <head>
4 
+        <meta charset="UTF-8">
5 
+        <title>Crossword Extractor</title>
6 
+    </head>
7 
+    <body>
8 
+        <form enctype="multipart/form-data" method="post" action="">
9 
+            <div>{{ form }}</div>
10 
+            <div><input type="submit" value="Upload"></div>
11 
+        </form>
12 
+    </body>
13 
+</html>
14 
+<!--
15 
+vim:ts=4:sw=4:expandtab
16 
+-->

+ 33 - 0
home/jinja2/home/output.html
查看文件 

@@ -0,0 +1,33 @@
1 
+<!DOCTYPE html>
2 
+<html>
3 
+    <head>
4 
+        <meta charset="UTF-8">
5 
+        <title>Extracted Crossword</title>
6 
+        <style>
7 
+            img {
8 
+                width: 600px;
9 
+            }
10 
+        </style>
11 
+    </head>
12 
+    <body>
13 
+        {% if warnings %}
14 
+        <div>
15 
+            Warnings:
16 
+            <ul>
17 
+                {% for warning in warnings %}
18 
+                    <li>{{ warning }}</li>
19 
+                {% endfor %}
20 
+            </ul>
21 
+        </div>
22 
+        {% endif %}
23 
+        <div>
24 
+            <a href="{{ image_url }}" download="{{ image_file_name }}">Download</a>
25 
+        </div>
26 
+        <div>
27 
+            <img alt="Crossword" src="{{ image_url }}">
28 
+        </div>
29 
+    </body>
30 
+</html>
31 
+<!--
32 
+vim:ts=4:sw=4:expandtab
33 
+-->

+ 0 - 0
home/management/__init__.py
查看文件


+ 0 - 0
home/management/commands/__init__.py
查看文件


+ 104 - 0
home/management/commands/extract.py
查看文件 

@@ -0,0 +1,104 @@
1 
+import cv2
2 
+from django.core.management.base import BaseCommand
3 
+from django.template.loader import render_to_string
4 
+from ...xword import extract_crossword_grid, draw_grid
5 
+
6 
+
7 
+class Command(BaseCommand):
8 
+    help = "Extracts a clean crossword image from a photograph."
9 
+
10 
+    def add_arguments(self, parser):
11 
+        parser.add_argument('input_file_name')
12 
+
13 
+        parser.add_argument('--debug', action='store_true')
14 
+
15 
+        parser.add_argument('--remove-colours', action='store_true')
16 
+        parser.add_argument('--colour-removal-threshold', type=int, default=48)
17 
+
18 
+        parser.add_argument('--gaussian-blur-size', type=int, default=11)
19 
+        parser.add_argument('--adaptive-threshold-block-size', type=int, default=11)
20 
+        parser.add_argument('--adaptive-threshold-mean-adjustment', type=int, default=2)
21 
+
22 
+        parser.add_argument('--not-square', action='store_true')
23 
+
24 
+        parser.add_argument('--num-dilations', type=int, default=1)
25 
+
26 
+        parser.add_argument('--contour-erosion-kernel-size', type=int, default=5)
27 
+        parser.add_argument('--contour-erosion-iterations', type=int, default=5)
28 
+        parser.add_argument('--line-detector-element-size', type=int, default=51)
29 
+
30 
+        parser.add_argument('--sampling-block-size-ratio', type=float, default=0.25)
31 
+        parser.add_argument('--sampling-threshold-quantile', type=float, default=0.3)
32 
+        parser.add_argument('--sampling-threshold', type=int)
33 
+
34 
+        parser.add_argument('--grid-line-thickness', type=int, default=4)
35 
+        parser.add_argument('--grid-square-size', type=int, default=64)
36 
+        parser.add_argument('--grid-border-size', type=int, default=20)
37 
+
38 
+        group = parser.add_mutually_exclusive_group()
39 
+        group.add_argument('--out')
40 
+        group.add_argument('--html')
41 
+
42 
+    def handle(self, *args, **options):
43 
+        warnings, grid, num_rows, num_cols, block_img = extract_crossword_grid(
44 
+            options['input_file_name'],
45 
+            callback=debug_callback if options['debug'] else None,
46 
+            remove_colours=options['remove_colours'],
47 
+            colour_removal_threshold=options['colour_removal_threshold'],
48 
+            gaussian_blur_size=options['gaussian_blur_size'],
49 
+            adaptive_threshold_block_size=options['adaptive_threshold_block_size'],
50 
+            adaptive_threshold_mean_adjustment=options['adaptive_threshold_mean_adjustment'],
51 
+            square=not options['not_square'],
52 
+            num_dilations=options['num_dilations'],
53 
+            contour_erosion_kernel_size=options['contour_erosion_kernel_size'],
54 
+            contour_erosion_iterations=options['contour_erosion_iterations'],
55 
+            line_detector_element_size=options['line_detector_element_size'],
56 
+            sampling_block_size_ratio=options['sampling_block_size_ratio'],
57 
+            sampling_threshold_quantile=options['sampling_threshold_quantile'],
58 
+            sampling_threshold=options['sampling_threshold']
59 
+        )
60 
+
61 
+        if options['debug']:
62 
+            debug_callback('square', block_img)
63 
+
64 
+        for warning in warnings:
65 
+            print('WARNING: ' + warning)
66 
+
67 
+        if options['html'] is not None:
68 
+            html = render_to_string('home/grid.html', {
69 
+                'grid': grid,
70 
+                'num_rows': num_rows,
71 
+                'num_cols': num_cols,
72 
+            })
73 
+
74 
+            with open(options['html'], 'w') as f:
75 
+                f.write(html)
76 
+
77 
+        else:
78 
+            image = draw_grid(
79 
+                grid,
80 
+                num_rows,
81 
+                num_cols,
82 
+                grid_line_thickness=options['grid_line_thickness'],
83 
+                grid_square_size=options['grid_square_size'],
84 
+                grid_border_size=options['grid_border_size']
85 
+            )
86 
+
87 
+            if options['debug'] or (options['out'] is None):
88 
+                debug_callback('output', image)
89 
+
90 
+            if options['out'] is not None:
91 
+                _, png = cv2.imencode('.png', image)
92 
+
93 
+                with open(options['out'], 'wb') as f:
94 
+                    f.write(png.tobytes())
95 
+
96 
+        if options['debug'] or (options['out'] is None):
97 
+            while cv2.waitKey() & 0xFF != ord('q'):
98 
+                pass
99 
+            cv2.destroyAllWindows()
100 
+
101 
+
102 
+def debug_callback(name, image):
103 
+    cv2.namedWindow(name, cv2.WINDOW_NORMAL)
104 
+    cv2.imshow(name, image)

+ 0 - 16
home/templates/home/index.html
查看文件 

@@ -1,16 +0,0 @@
1 
-<!DOCTYPE html>
2 
-<html>
3 
-  <head>
4 
-    <meta charset="UTF-8">
5 
-    <title>Crossword Extractor</title>
6 
-  </head>
7 
-  <body>
8 
-    <form enctype="multipart/form-data" method="post" action="">
9 
-      <div>{{ form }}</div>
10 
-      <div><input type="submit" value="Upload"></div>
11 
-    </form>
12 
-  </body>
13 
-</html>
14 
-<!--
15 
-vim:ts=2:sw=2:expandtab
16 
--->

+ 0 - 33
home/templates/home/output.html
查看文件 

@@ -1,33 +0,0 @@
1 
-<!DOCTYPE html>
2 
-<html>
3 
-  <head>
4 
-    <meta charset="UTF-8">
5 
-    <title>Extracted Crossword</title>
6 
-    <style>
7 
-      img {
8 
-        width: 600px;
9 
-      }
10 
-    </style>
11 
-  </head>
12 
-  <body>
13 
-    {% if warnings %}
14 
-    <div>
15 
-      Warnings:
16 
-      <ul>
17 
-        {% for warning in warnings %}
18 
-          <li>{{ warning }}</li>
19 
-        {% endfor %}
20 
-      </ul>
21 
-    </div>
22 
-    {% endif %}
23 
-    <div>
24 
-      <a href="{{ image_url }}" download="{{ image_file_name }}">Download</a>
25 
-    </div>
26 
-    <div>
27 
-      <img alt="Crossword" src="{{ image_url }}">
28 
-    </div>
29 
-  </body>
30 
-</html>
31 
-<!--
32 
-vim:ts=2:sw=2:expandtab
33 
--->

+ 9 - 3
home/views.py
查看文件 

@@ -1,10 +1,11 @@
1 1 
 from base64 import b64encode
2 
+import cv2
2 3 
 from datetime import datetime
3 4 
 from django.http import HttpResponse, HttpResponseBadRequest
4 5 
 from django.shortcuts import render
5 6 
 from django.views import View
6 7 
 from .forms import CrosswordForm
7 
-from .xword import extract_crossword
8 
+from .xword import extract_crossword_grid, draw_grid
8 9
9 10
10 11 
 class HomeView(View):
 
@@ -15,9 +16,14 @@ class HomeView(View):
15 16 
         form = CrosswordForm(request.POST, request.FILES)
16 17 
         if not form.is_valid():
17 18 
             return HttpResponseBadRequest('Invalid form data')
18 
-        image, warnings = extract_crossword(form.cleaned_data['file'].temporary_file_path())
19 
+        warnings, grid, num_rows, num_cols, block_img = extract_crossword_grid(
20 
+            form.cleaned_data['file'].temporary_file_path(),
21 
+            remove_colours=False
22 
+        )
23 
+        image = draw_grid(grid, num_rows, num_cols)
24 
+        _, png = cv2.imencode('.png', image)
19 25 
         return render(request, 'home/output.html', {
20 26 
             'warnings': warnings,
21 27 
             'image_file_name': 'xword_{}.png'.format(datetime.now().strftime('%Y%m%d_%H%M%S')),
22 
-            'image_url': 'data:image/png;base64,' + b64encode(image).decode()
28 
+            'image_url': 'data:image/png;base64,' + b64encode(png.tobytes()).decode()
23 29 
         })

+ 67 - 27
home/xword.py
查看文件 

@@ -5,6 +5,25 @@ import copy
5 5 
 import argparse
6 6
7 7
8 
+def non_greys_to_white(img, threshold=48):
9 
+    b, g, r = cv2.split(img)
10 
+    rgb_diff = cv2.subtract(cv2.max(cv2.max(b, g), r), cv2.min(cv2.min(b, g), r))
11 
+    filtered = img.copy()
12 
+    filtered[np.where(rgb_diff > threshold)] = (255, 255, 255)
13 
+    return filtered
14 
+
15 
+
16 
+def load_image_as_greyscale(file_name, filter_colours, colour_filter_threshold):
17 
+    img = cv2.imread(file_name)
18 
+    if img is None:
19 
+        raise RuntimeError("Failed to load image")
20 
+
21 
+    if filter_colours:
22 
+        img = non_greys_to_white(img, colour_filter_threshold)
23 
+
24 
+    return cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
25 
+
26 
+
8 27 
 def preprocess_image(original, gaussian_blur_size, adaptive_threshold_block_size, adaptive_threshold_mean_adjustment, num_dilations):
9 28 
     img = cv2.GaussianBlur(original, (gaussian_blur_size, gaussian_blur_size), 0)
10 29 
     img = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, adaptive_threshold_block_size, adaptive_threshold_mean_adjustment)
 
@@ -136,25 +155,41 @@ def extract_grid_colours(img, num_rows, num_cols, sampling_block_size_ratio):
136 155 
 def grid_colours_to_blocks(grid_colours, num_rows, num_cols, sampling_threshold):
137 156 
     grid = copy.deepcopy(grid_colours)
138 157 
     warning = False
139 
-    for row in range(round(num_rows / 2)):
158 
+
159 
+    midpoint = num_rows // 2 + (0 if num_rows % 2 == 0 else 1)
160 
+    for row in range(midpoint):
140 161 
         for col in range(num_cols):
162 
+            # If there is an odd number of rows then row and row2 will point to
163 
+            # the same row when we reach the middle. Doesn't seem worth adding a
164 
+            # special case.
141 165 
             row2 = num_rows - row - 1
142 166 
             col2 = num_cols - col - 1
143 167 
             delta1 = grid_colours[row][col] - sampling_threshold
144 168 
             delta2 = grid_colours[row2][col2] - sampling_threshold
145 169
146 170 
             if (delta1 > 0) and (delta2 > 0):
147 
-                block = 0
171 
+                filled = False
148 172 
             elif (delta1 < 0) and (delta2 < 0):
149 
-                block = 1
173 
+                filled = True
150 174 
             else:
151 175 
                 warning = True
152 176 
                 if abs(delta1) > abs(delta2):
153 
-                    block = 1 if delta1 < 0 else 0
177 
+                    filled = delta1 < 0
154 178 
                 else:
155 
-                    block = 1 if delta2 < 0 else 0
179 
+                    filled = delta2 < 0
180 
+
181 
+            grid[row][col] = {'filled': filled}
182 
+            grid[row2][col2] = {'filled': filled}
156 183
157 
-            grid[row][col] = grid[row2][col2] = block
184 
+    number = 1
185 
+    for row in range(num_rows):
186 
+        for col in range(num_cols):
187 
+            if (not grid[row][col]['filled'] and (
188 
+                (((col == 0) or grid[row][col - 1]['filled']) and (col < num_cols - 1) and not grid[row][col + 1]['filled']) or
189 
+                (((row == 0) or grid[row - 1][col]['filled']) and (row < num_rows - 1) and not grid[row + 1][col]['filled'])
190 
+            )):
191 
+                grid[row][col]['number'] = number
192 
+                number += 1
158 193
159 194 
     return warning, grid
160 195 
 
@@ -169,38 +204,32 @@ def draw_point(image, point, colour):
169 204 
                 image[y, x] = colour
170 205
171 206
172 
-def show_image(image):
173 
-    cv2.namedWindow('xword', cv2.WINDOW_NORMAL)
174 
-    cv2.imshow('xword', image)
175 
-    while cv2.waitKey() & 0xFF != ord('q'):
176 
-        pass
177 
-    cv2.destroyAllWindows()
178 
-
179 
-
180 
-def extract_crossword(
207 
+def extract_crossword_grid(
181 208 
     file_name,
209 
+    callback=None,
210 
+    remove_colours=False,
211 
+    colour_removal_threshold=48,
182 212 
     gaussian_blur_size=11,
183 213 
     adaptive_threshold_block_size=11,
184 214 
     adaptive_threshold_mean_adjustment=2,
185 215 
     square=True,
186 216 
     num_dilations=1,
187 217 
     contour_erosion_kernel_size=5,
188 
-    contour_erosion_iterations=6,
218 
+    contour_erosion_iterations=5,
189 219 
     line_detector_element_size=51,
190 220 
     sampling_block_size_ratio=0.25,
191 221 
     sampling_threshold_quantile=0.3,
192 
-    sampling_threshold=None,
193 
-    grid_line_thickness=4,
194 
-    grid_square_size=64,
195 
-    grid_border_size=20,
222 
+    sampling_threshold=None
196 223 
 ):
197 224 
     warnings = []
198 225
199 
-    original = cv2.imread(file_name, cv2.IMREAD_GRAYSCALE)
200 
-    if original is None:
201 
-        raise RuntimeError("Failed to load image")
226 
+    original = load_image_as_greyscale(file_name, remove_colours, colour_removal_threshold)
227 
+    if callback is not None:
228 
+        callback('original', original)
202 229
203 230 
     img = preprocess_image(original, gaussian_blur_size, adaptive_threshold_block_size, adaptive_threshold_mean_adjustment, num_dilations)
231 
+    if callback is not None:
232 
+        callback('preprocessed', img)
204 233
205 234 
     biggest = find_biggest_contour(img)
206 235 
     biggest = erode_contour(img.shape, biggest, contour_erosion_kernel_size, contour_erosion_iterations)
 
@@ -208,6 +237,8 @@ def extract_crossword(
208 237 
     top_left, top_right, bottom_right, bottom_left = get_contour_corners(img, biggest)
209 238
210 239 
     img = extract_square(img, top_left, top_right, bottom_right, bottom_left)
240 
+    if callback is not None:
241 
+        callback('pre-fft', img)
211 242
212 243 
     num_rows = get_line_frequency(img, line_detector_element_size, 1)
213 244 
     num_cols = get_line_frequency(img, line_detector_element_size, 0)
 
@@ -226,6 +257,17 @@ def extract_crossword(
226 257 
     if warning:
227 258 
         warnings.append("Some blocks may be the wrong colour")
228 259
260 
+    return warnings, grid, num_rows, num_cols, block_img
261 
+
262 
+
263 
+def draw_grid(
264 
+    grid,
265 
+    num_rows,
266 
+    num_cols,
267 
+    grid_line_thickness=4,
268 
+    grid_square_size=64,
269 
+    grid_border_size=20
270 
+):
229 271 
     step = grid_square_size + grid_line_thickness
230 272 
     grid_height = num_rows * step + grid_line_thickness
231 273 
     grid_width = num_cols * step + grid_line_thickness
 
@@ -234,9 +276,7 @@ def extract_crossword(
234 276 
     for row in range(num_rows):
235 277 
         y = row * step + grid_line_thickness + grid_border_size
236 278 
         for col in range(num_cols):
237 
-            if grid[row][col] == 0:
279 
+            if not grid[row][col]['filled']:
238 280 
                 x = col * step + grid_line_thickness + grid_border_size
239 281 
                 cv2.rectangle(output, (x, y), (x + grid_square_size - 1, y + grid_square_size - 1), 255, -1)
240 
-
241 
-    _, png = cv2.imencode('.png', output)
242 
-    return png.tobytes(), warnings
282 
+    return output

+ 2 - 0
requirements.in
查看文件 

@@ -1,4 +1,6 @@
1 1 
 chaussette
2 2 
 django
3 
+django-extensions
4 
+jinja2
3 5 
 opencv-python
4 6 
 pip-tools

+ 4 - 1
requirements.txt
查看文件 

@@ -7,10 +7,13 @@
7 7 
 asgiref==3.2.3            # via django
8 8 
 chaussette==1.3.0
9 9 
 click==7.0                # via pip-tools
10 
+django-extensions==2.2.6
10 11 
 django==3.0.2
12 
+jinja2==2.11.0
13 
+markupsafe==1.1.1         # via jinja2
11 14 
 numpy==1.18.1             # via opencv-python
12 15 
 opencv-python==4.1.2.30
13 16 
 pip-tools==4.4.0
14 17 
 pytz==2019.3              # via django
15 
-six==1.14.0               # via chaussette, pip-tools
18 
+six==1.14.0               # via chaussette, django-extensions, pip-tools
16 19 
 sqlparse==0.3.0           # via django

+ 12 - 0
xword/jinja2.py
查看文件 

@@ -0,0 +1,12 @@
1 
+#from django.contrib.staticfiles.storage import staticfiles_storage
2 
+#from django.urls import reverse
3 
+from jinja2 import Environment
4 
+
5 
+
6 
+def environment(**options):
7 
+    env = Environment(**options)
8 
+    env.globals.update({
9 
+        #‘static’: staticfiles_storage.url,
10 
+        #‘url’: reverse,
11 
+    })
12 
+    return env

+ 3 - 1
xword/settings.py
查看文件 

@@ -11,6 +11,7 @@ DEBUG = False
11 11 
 ALLOWED_HOSTS = OVERRIDE_ME
12 12
13 13 
 INSTALLED_APPS = [
14 
+    'django_extensions',
14 15 
     'home.apps.HomeConfig',
15 16 
 ]
16 17 
 
@@ -24,10 +25,11 @@ ROOT_URLCONF = 'xword.urls'
24 25
25 26 
 TEMPLATES = [
26 27 
     {
27 
-        'BACKEND': 'django.template.backends.django.DjangoTemplates',
28 
+        'BACKEND': 'django.template.backends.jinja2.Jinja2',
28 29 
         'DIRS': [],
29 30 
         'APP_DIRS': True,
30 31 
         'OPTIONS': {
32 
+            'environment': 'xword.jinja2.environment',
31 33 
             'context_processors': [
32 34 
                 'django.template.context_processors.debug',
33 35 
                 'django.template.context_processors.request',