개정판 4315c4e5
issue #663: craft testing
Change-Id: Ibb30cfd290868653d28335cfbd7f1ea388aa86df
| DTI_PID/DTI_PID/TextDetector.py | ||
|---|---|---|
| 114 | 114 |
#from imutils.object_detection import non_max_suppression |
| 115 | 115 |
from AppDocData import AppDocData |
| 116 | 116 |
|
| 117 |
list = [] |
|
| 117 |
res_list = []
|
|
| 118 | 118 |
ocr_image = None |
| 119 | 119 |
try: |
| 120 | 120 |
app_doc_data = AppDocData.instance() |
| ... | ... | |
| 131 | 131 |
not_containing_bbox, binary_image = self.getTextBox(ocr_image, imgGray, maxTextSize, minSize) |
| 132 | 132 |
|
| 133 | 133 |
# using craft |
| 134 |
#not_containing_bbox, binary_image = self.getTextBox_craft(ocr_image, maxTextSize, minSize)
|
|
| 134 |
#return self.getTextBox_craft(ocr_image, maxTextSize, minSize, offset_x, offset_y)
|
|
| 135 | 135 |
|
| 136 | 136 |
rects = [] |
| 137 | 137 |
|
| ... | ... | |
| 198 | 198 |
# if there is no boxes which contains |
| 199 | 199 |
if not matches: |
| 200 | 200 |
angle = rect[0] |
| 201 |
list.append(ti.TextInfo('', round(offset_x) + rect[1].x(), round(offset_y) + rect[1].y(), rect[1].width(),
|
|
| 201 |
res_list.append(ti.TextInfo('', round(offset_x) + rect[1].x(), round(offset_y) + rect[1].y(), rect[1].width(),
|
|
| 202 | 202 |
rect[1].height(), angle)) |
| 203 | 203 |
except Exception as ex: |
| 204 | 204 |
message = 'error occurred({}) in {}:{}'.format(repr(ex), sys.exc_info()[-1].tb_frame.f_code.co_filename,
|
| 205 | 205 |
sys.exc_info()[-1].tb_lineno) |
| 206 | 206 |
print(message) |
| 207 | 207 |
|
| 208 |
return list, ocr_image |
|
| 208 |
return res_list, ocr_image |
|
| 209 |
|
|
| 210 |
''' |
|
| 211 |
def getTextBox_craft(self, ocr_image, maxTextSize, minSize, offset_x, offset_y): |
|
| 212 |
""" get text box by using craft """ |
|
| 209 | 213 |
|
| 210 |
def getTextBox_craft(self, ocr_image, maxTextSize, minSize): |
|
| 211 |
pass |
|
| 212 |
''' |
|
| 213 | 214 |
from CRAFT_pytorch_master import text_craft |
| 214 | 215 |
from AppDocData import AppDocData |
| 215 | 216 |
|
| ... | ... | |
| 227 | 228 |
for box in boxes: |
| 228 | 229 |
rects.append(QRect(box[0], box[1], box[4] - box[0], box[5] - box[1])) |
| 229 | 230 |
|
| 230 |
return rects, binary_image |
|
| 231 |
''' |
|
| 231 |
#configs = app_doc_data.getConfigs('Text Recognition', 'Merge Size')
|
|
| 232 |
#mergeSize = int(configs[0].value) if 1 == len(configs) else 10 |
|
| 233 |
#gap_size = mergeSize / 2 |
|
| 234 |
gap_size = 3 |
|
| 235 |
|
|
| 236 |
verticals = [] |
|
| 237 |
horizontals = [] |
|
| 238 |
for rect in rects: |
|
| 239 |
if rect.width() < minSize and rect.height() < maxTextSize: |
|
| 240 |
rect._vertical = False |
|
| 241 |
horizontals.append(rect) |
|
| 242 |
elif rect.height() < minSize and rect.width() < maxTextSize: |
|
| 243 |
rect._vertical = True |
|
| 244 |
verticals.append(rect) |
|
| 245 |
elif rect.width() < minSize or rect.height() < minSize: |
|
| 246 |
continue |
|
| 247 |
elif rect.height() > rect.width(): |
|
| 248 |
rect._vertical = True |
|
| 249 |
verticals.append(rect) |
|
| 250 |
else: |
|
| 251 |
rect._vertical = False |
|
| 252 |
horizontals.append(rect) |
|
| 253 |
|
|
| 254 |
v_merges = [] |
|
| 255 |
for vertical1 in verticals: |
|
| 256 |
for vertical2 in verticals: |
|
| 257 |
if vertical1 is vertical2: |
|
| 258 |
continue |
|
| 259 |
if abs(vertical1.center().x() - vertical2.center().x()) < gap_size: |
|
| 260 |
t1, t2 = vertical1.top() + gap_size, vertical2.top() + gap_size |
|
| 261 |
b1, b2 = vertical1.bottom() - gap_size, vertical2.bottom() - gap_size |
|
| 262 |
if not (b2 < t1 or t2 < b1): |
|
| 263 |
inserted = False |
|
| 264 |
for merge in v_merges: |
|
| 265 |
if vertical1 in merge and vertical2 in merge: |
|
| 266 |
inserted = True |
|
| 267 |
break |
|
| 268 |
elif vertical1 in merge and vertical2 not in merge: |
|
| 269 |
merge.append(vertical2) |
|
| 270 |
inserted = True |
|
| 271 |
break |
|
| 272 |
elif vertical2 in merge and vertical1 not in merge: |
|
| 273 |
merge.append(vertical1) |
|
| 274 |
inserted = True |
|
| 275 |
break |
|
| 276 |
if not inserted: |
|
| 277 |
v_merges.append([vertical1, vertical2]) |
|
| 278 |
|
|
| 279 |
h_merges = [] |
|
| 280 |
for horizontal1 in horizontals: |
|
| 281 |
for horizontal2 in horizontals: |
|
| 282 |
if horizontal1 is horizontal2: |
|
| 283 |
continue |
|
| 284 |
if abs(horizontal1.center().y() - horizontal2.center().y()) < gap_size: |
|
| 285 |
l1, l2 = horizontal1.left() - gap_size, horizontal2.left() - gap_size |
|
| 286 |
r1, r2 = horizontal1.right() + gap_size, horizontal2.right() + gap_size |
|
| 287 |
if not (l2 < r1 or r2 < l1): |
|
| 288 |
inserted = False |
|
| 289 |
for merge in h_merges: |
|
| 290 |
if horizontal1 in merge and horizontal2 in merge: |
|
| 291 |
inserted = True |
|
| 292 |
break |
|
| 293 |
elif horizontal1 in merge and horizontal2 not in merge: |
|
| 294 |
merge.append(horizontal2) |
|
| 295 |
inserted = True |
|
| 296 |
break |
|
| 297 |
elif horizontal2 in merge and horizontal1 not in merge: |
|
| 298 |
merge.append(horizontal1) |
|
| 299 |
inserted = True |
|
| 300 |
break |
|
| 301 |
if not inserted: |
|
| 302 |
h_merges.append([horizontal1, horizontal2]) |
|
| 303 |
|
|
| 304 |
for merge in v_merges + h_merges: |
|
| 305 |
for rect in merge: |
|
| 306 |
if rect in merge: |
|
| 307 |
rects.remove(rect) |
|
| 308 |
else: |
|
| 309 |
print(str(rect)) |
|
| 310 |
|
|
| 311 |
for merge in v_merges: |
|
| 312 |
max_x, max_y, min_x, min_y = 0, 0, sys.maxsize, sys.maxsize |
|
| 313 |
for rect in merge: |
|
| 314 |
if rect.left() < min_x: |
|
| 315 |
min_x = rect.left() |
|
| 316 |
if rect.right() > max_x: |
|
| 317 |
max_x = rect.right() |
|
| 318 |
if rect.top() > max_y: |
|
| 319 |
max_y = rect.top() |
|
| 320 |
if rect.bottom() < min_y: |
|
| 321 |
min_y = rect.bottom() |
|
| 322 |
|
|
| 323 |
rect = QRect(min_x, max_y, max_x - min_x, max_y - min_x) |
|
| 324 |
rect._vertical = True |
|
| 325 |
rects.append(rect) |
|
| 326 |
|
|
| 327 |
for merge in h_merges: |
|
| 328 |
max_x, max_y, min_x, min_y = 0, 0, sys.maxsize, sys.maxsize |
|
| 329 |
for rect in merge: |
|
| 330 |
if rect.left() < min_x: |
|
| 331 |
min_x = rect.left() |
|
| 332 |
if rect.right() > max_x: |
|
| 333 |
max_x = rect.right() |
|
| 334 |
if rect.top() > max_y: |
|
| 335 |
max_y = rect.top() |
|
| 336 |
if rect.bottom() < min_y: |
|
| 337 |
min_y = rect.bottom() |
|
| 338 |
|
|
| 339 |
rect = QRect(min_x, max_y, max_x - min_x, max_y - min_x) |
|
| 340 |
rect._vertical = False |
|
| 341 |
rects.append(rect) |
|
| 342 |
|
|
| 343 |
res_rects = [] |
|
| 344 |
for rect in rects: |
|
| 345 |
res_rects.append(ti.TextInfo('', round(offset_x) + rect.x(), round(offset_y) + rect.y(), rect.width(),
|
|
| 346 |
rect.height(), 90 if rect._vertical else 0)) |
|
| 347 |
|
|
| 348 |
return res_rects, binary_image |
|
| 349 |
''' |
|
| 232 | 350 |
|
| 233 | 351 |
def getTextBox(self, ocr_image, imgGray, maxTextSize, minSize): |
| 234 | 352 |
""" get text box """ |
내보내기 Unified diff
