utility.py 19.6 KB
Newer Older
LDOUBLEV's avatar
LDOUBLEV committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import argparse
WenmuZhou's avatar
WenmuZhou committed
16
import os
WenmuZhou's avatar
WenmuZhou committed
17
import sys
LDOUBLEV's avatar
LDOUBLEV committed
18
19
import cv2
import numpy as np
LDOUBLEV's avatar
LDOUBLEV committed
20
21
import json
from PIL import Image, ImageDraw, ImageFont
22
import math
WenmuZhou's avatar
WenmuZhou committed
23
from paddle import inference
24
25
import time
from ppocr.utils.logging import get_logger
WenmuZhou's avatar
WenmuZhou committed
26

27
logger = get_logger()
LDOUBLEV's avatar
LDOUBLEV committed
28
29


30
31
def str2bool(v):
    return v.lower() in ("true", "t", "1")
LDOUBLEV's avatar
LDOUBLEV committed
32
33


WenmuZhou's avatar
WenmuZhou committed
34
35
def init_args():
    parser = argparse.ArgumentParser()
36
    # params for prediction engine
WenmuZhou's avatar
WenmuZhou committed
37
38
39
40
41
42
    parser.add_argument("--use_gpu", type=str2bool, default=True)
    parser.add_argument("--ir_optim", type=str2bool, default=True)
    parser.add_argument("--use_tensorrt", type=str2bool, default=False)
    parser.add_argument("--use_fp16", type=str2bool, default=False)
    parser.add_argument("--gpu_mem", type=int, default=500)

WenmuZhou's avatar
WenmuZhou committed
43
    # params for text detector
WenmuZhou's avatar
WenmuZhou committed
44
45
46
47
48
49
    parser.add_argument("--image_dir", type=str)
    parser.add_argument("--det_algorithm", type=str, default='DB')
    parser.add_argument("--det_model_dir", type=str)
    parser.add_argument("--det_limit_side_len", type=float, default=960)
    parser.add_argument("--det_limit_type", type=str, default='max')

WenmuZhou's avatar
WenmuZhou committed
50
    # DB parmas
WenmuZhou's avatar
WenmuZhou committed
51
52
53
54
55
56
    parser.add_argument("--det_db_thresh", type=float, default=0.3)
    parser.add_argument("--det_db_box_thresh", type=float, default=0.5)
    parser.add_argument("--det_db_unclip_ratio", type=float, default=1.6)
    parser.add_argument("--max_batch_size", type=int, default=10)
    parser.add_argument("--use_dilation", type=bool, default=False)
    parser.add_argument("--det_db_score_mode", type=str, default="fast")
WenmuZhou's avatar
WenmuZhou committed
57
    # EAST parmas
WenmuZhou's avatar
WenmuZhou committed
58
59
60
61
    parser.add_argument("--det_east_score_thresh", type=float, default=0.8)
    parser.add_argument("--det_east_cover_thresh", type=float, default=0.1)
    parser.add_argument("--det_east_nms_thresh", type=float, default=0.2)

WenmuZhou's avatar
WenmuZhou committed
62
    # SAST parmas
WenmuZhou's avatar
WenmuZhou committed
63
64
65
66
    parser.add_argument("--det_sast_score_thresh", type=float, default=0.5)
    parser.add_argument("--det_sast_nms_thresh", type=float, default=0.2)
    parser.add_argument("--det_sast_polygon", type=bool, default=False)

WenmuZhou's avatar
WenmuZhou committed
67
    # params for text recognizer
WenmuZhou's avatar
WenmuZhou committed
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
    parser.add_argument("--rec_algorithm", type=str, default='CRNN')
    parser.add_argument("--rec_model_dir", type=str)
    parser.add_argument("--rec_image_shape", type=str, default="3, 32, 320")
    parser.add_argument("--rec_char_type", type=str, default='ch')
    parser.add_argument("--rec_batch_num", type=int, default=6)
    parser.add_argument("--max_text_length", type=int, default=25)
    parser.add_argument(
        "--rec_char_dict_path",
        type=str,
        default="./ppocr/utils/ppocr_keys_v1.txt")
    parser.add_argument("--use_space_char", type=str2bool, default=True)
    parser.add_argument(
        "--vis_font_path", type=str, default="./doc/fonts/simfang.ttf")
    parser.add_argument("--drop_score", type=float, default=0.5)

Jethong's avatar
Jethong committed
83
    # params for e2e
WenmuZhou's avatar
WenmuZhou committed
84
85
86
87
88
    parser.add_argument("--e2e_algorithm", type=str, default='PGNet')
    parser.add_argument("--e2e_model_dir", type=str)
    parser.add_argument("--e2e_limit_side_len", type=float, default=768)
    parser.add_argument("--e2e_limit_type", type=str, default='max')

Jethong's avatar
Jethong committed
89
    # PGNet parmas
WenmuZhou's avatar
WenmuZhou committed
90
91
92
93
94
95
96
    parser.add_argument("--e2e_pgnet_score_thresh", type=float, default=0.5)
    parser.add_argument(
        "--e2e_char_dict_path", type=str, default="./ppocr/utils/ic15_dict.txt")
    parser.add_argument("--e2e_pgnet_valid_set", type=str, default='totaltext')
    parser.add_argument("--e2e_pgnet_polygon", type=bool, default=True)
    parser.add_argument("--e2e_pgnet_mode", type=str, default='fast')

WenmuZhou's avatar
WenmuZhou committed
97
    # params for text classifier
WenmuZhou's avatar
WenmuZhou committed
98
99
100
101
102
103
104
105
106
107
108
109
110
111
    parser.add_argument("--use_angle_cls", type=str2bool, default=False)
    parser.add_argument("--cls_model_dir", type=str)
    parser.add_argument("--cls_image_shape", type=str, default="3, 48, 192")
    parser.add_argument("--label_list", type=list, default=['0', '180'])
    parser.add_argument("--cls_batch_num", type=int, default=6)
    parser.add_argument("--cls_thresh", type=float, default=0.9)

    parser.add_argument("--enable_mkldnn", type=str2bool, default=False)
    parser.add_argument("--cpu_threads", type=int, default=10)
    parser.add_argument("--use_pdserving", type=str2bool, default=False)

    parser.add_argument("--use_mp", type=str2bool, default=False)
    parser.add_argument("--total_process_num", type=int, default=1)
    parser.add_argument("--process_id", type=int, default=0)
WenmuZhou's avatar
WenmuZhou committed
112
    parser.add_argument("--show_log", type=str2bool, default=True)
WenmuZhou's avatar
WenmuZhou committed
113
    return parser
WenmuZhou's avatar
WenmuZhou committed
114

115

116
def parse_args():
WenmuZhou's avatar
WenmuZhou committed
117
    parser = init_args()
LDOUBLEV's avatar
LDOUBLEV committed
118
119
120
    return parser.parse_args()


WenmuZhou's avatar
WenmuZhou committed
121
122
123
124
125
def create_predictor(args, mode, logger):
    if mode == "det":
        model_dir = args.det_model_dir
    elif mode == 'cls':
        model_dir = args.cls_model_dir
Jethong's avatar
Jethong committed
126
    elif mode == 'rec':
WenmuZhou's avatar
WenmuZhou committed
127
        model_dir = args.rec_model_dir
WenmuZhou's avatar
WenmuZhou committed
128
129
    elif mode == 'structure':
        model_dir = args.structure_model_dir
Jethong's avatar
Jethong committed
130
131
    else:
        model_dir = args.e2e_model_dir
WenmuZhou's avatar
WenmuZhou committed
132
133
134
135

    if model_dir is None:
        logger.info("not find {} model file path {}".format(mode, model_dir))
        sys.exit(0)
WenmuZhou's avatar
WenmuZhou committed
136
137
    model_file_path = model_dir + "/inference.pdmodel"
    params_file_path = model_dir + "/inference.pdiparams"
WenmuZhou's avatar
WenmuZhou committed
138
139
140
141
142
143
144
    if not os.path.exists(model_file_path):
        logger.info("not find model file path {}".format(model_file_path))
        sys.exit(0)
    if not os.path.exists(params_file_path):
        logger.info("not find params file path {}".format(params_file_path))
        sys.exit(0)

WenmuZhou's avatar
WenmuZhou committed
145
    config = inference.Config(model_file_path, params_file_path)
WenmuZhou's avatar
WenmuZhou committed
146
147
148

    if args.use_gpu:
        config.enable_use_gpu(args.gpu_mem, 0)
LDOUBLEV's avatar
LDOUBLEV committed
149
150
        if args.use_tensorrt:
            config.enable_tensorrt_engine(
LDOUBLEV's avatar
LDOUBLEV committed
151
152
                precision_mode=inference.PrecisionType.Float32,
                max_batch_size=args.max_batch_size,
WenmuZhou's avatar
WenmuZhou committed
153
                min_subgraph_size=3)  # skip the minmum trt subgraph
LDOUBLEV's avatar
LDOUBLEV committed
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
        if mode == "det" and "mobile" in model_file_path:
            min_input_shape = {
                "x": [1, 3, 50, 50],
                "conv2d_92.tmp_0": [1, 96, 20, 20],
                "conv2d_91.tmp_0": [1, 96, 10, 10],
                "nearest_interp_v2_1.tmp_0": [1, 96, 10, 10],
                "nearest_interp_v2_2.tmp_0": [1, 96, 20, 20],
                "nearest_interp_v2_3.tmp_0": [1, 24, 20, 20],
                "nearest_interp_v2_4.tmp_0": [1, 24, 20, 20],
                "nearest_interp_v2_5.tmp_0": [1, 24, 20, 20],
                "elementwise_add_7": [1, 56, 2, 2],
                "nearest_interp_v2_0.tmp_0": [1, 96, 2, 2]
            }
            max_input_shape = {
                "x": [1, 3, 2000, 2000],
                "conv2d_92.tmp_0": [1, 96, 400, 400],
                "conv2d_91.tmp_0": [1, 96, 200, 200],
                "nearest_interp_v2_1.tmp_0": [1, 96, 200, 200],
                "nearest_interp_v2_2.tmp_0": [1, 96, 400, 400],
                "nearest_interp_v2_3.tmp_0": [1, 24, 400, 400],
                "nearest_interp_v2_4.tmp_0": [1, 24, 400, 400],
                "nearest_interp_v2_5.tmp_0": [1, 24, 400, 400],
                "elementwise_add_7": [1, 56, 400, 400],
                "nearest_interp_v2_0.tmp_0": [1, 96, 400, 400]
            }
            opt_input_shape = {
                "x": [1, 3, 640, 640],
                "conv2d_92.tmp_0": [1, 96, 160, 160],
                "conv2d_91.tmp_0": [1, 96, 80, 80],
                "nearest_interp_v2_1.tmp_0": [1, 96, 80, 80],
                "nearest_interp_v2_2.tmp_0": [1, 96, 160, 160],
                "nearest_interp_v2_3.tmp_0": [1, 24, 160, 160],
                "nearest_interp_v2_4.tmp_0": [1, 24, 160, 160],
                "nearest_interp_v2_5.tmp_0": [1, 24, 160, 160],
                "elementwise_add_7": [1, 56, 40, 40],
                "nearest_interp_v2_0.tmp_0": [1, 96, 40, 40]
            }
        if mode == "det" and "server" in model_file_path:
            min_input_shape = {
                "x": [1, 3, 50, 50],
                "conv2d_59.tmp_0": [1, 96, 20, 20],
                "nearest_interp_v2_2.tmp_0": [1, 96, 20, 20],
                "nearest_interp_v2_3.tmp_0": [1, 24, 20, 20],
                "nearest_interp_v2_4.tmp_0": [1, 24, 20, 20],
                "nearest_interp_v2_5.tmp_0": [1, 24, 20, 20]
            }
            max_input_shape = {
                "x": [1, 3, 2000, 2000],
                "conv2d_59.tmp_0": [1, 96, 400, 400],
                "nearest_interp_v2_2.tmp_0": [1, 96, 400, 400],
                "nearest_interp_v2_3.tmp_0": [1, 24, 400, 400],
                "nearest_interp_v2_4.tmp_0": [1, 24, 400, 400],
                "nearest_interp_v2_5.tmp_0": [1, 24, 400, 400]
            }
            opt_input_shape = {
                "x": [1, 3, 640, 640],
                "conv2d_59.tmp_0": [1, 96, 160, 160],
                "nearest_interp_v2_2.tmp_0": [1, 96, 160, 160],
                "nearest_interp_v2_3.tmp_0": [1, 24, 160, 160],
                "nearest_interp_v2_4.tmp_0": [1, 24, 160, 160],
                "nearest_interp_v2_5.tmp_0": [1, 24, 160, 160]
            }
        elif mode == "rec":
            min_input_shape = {"x": [args.rec_batch_num, 3, 32, 10]}
            max_input_shape = {"x": [args.rec_batch_num, 3, 32, 2000]}
            opt_input_shape = {"x": [args.rec_batch_num, 3, 32, 320]}
        elif mode == "cls":
            min_input_shape = {"x": [args.rec_batch_num, 3, 48, 10]}
            max_input_shape = {"x": [args.rec_batch_num, 3, 48, 2000]}
            opt_input_shape = {"x": [args.rec_batch_num, 3, 48, 320]}
LDOUBLEV's avatar
LDOUBLEV committed
224
225
226
227
        else:
            min_input_shape = {"x": [1, 3, 10, 10]}
            max_input_shape = {"x": [1, 3, 1000, 1000]}
            opt_input_shape = {"x": [1, 3, 500, 500]}
LDOUBLEV's avatar
LDOUBLEV committed
228
229
230
        config.set_trt_dynamic_shape_info(min_input_shape, max_input_shape,
                                          opt_input_shape)

WenmuZhou's avatar
WenmuZhou committed
231
232
    else:
        config.disable_gpu()
233
234
235
        if hasattr(args, "cpu_threads"):
            config.set_cpu_math_library_num_threads(args.cpu_threads)
        else:
WenmuZhou's avatar
WenmuZhou committed
236
237
            # default cpu threads as 10
            config.set_cpu_math_library_num_threads(10)
WenmuZhou's avatar
WenmuZhou committed
238
239
240
241
242
        if args.enable_mkldnn:
            # cache 10 different shapes for mkldnn to avoid memory leak
            config.set_mkldnn_cache_capacity(10)
            config.enable_mkldnn()

LDOUBLEV's avatar
LDOUBLEV committed
243
244
    # enable memory optim
    config.enable_memory_optim()
WenmuZhou's avatar
WenmuZhou committed
245
246
    config.disable_glog_info()

WenmuZhou's avatar
WenmuZhou committed
247
248
    config.delete_pass("conv_transpose_eltwiseadd_bn_fuse_pass")
    config.switch_use_feed_fetch_ops(False)
WenmuZhou's avatar
WenmuZhou committed
249
250
    config.switch_ir_optim(True)
    if mode == 'structure':
WenmuZhou's avatar
WenmuZhou committed
251
        config.switch_ir_optim(False)
WenmuZhou's avatar
WenmuZhou committed
252
253
    # create predictor
    predictor = inference.create_predictor(config)
WenmuZhou's avatar
WenmuZhou committed
254
255
    input_names = predictor.get_input_names()
    for name in input_names:
WenmuZhou's avatar
WenmuZhou committed
256
        input_tensor = predictor.get_input_handle(name)
WenmuZhou's avatar
WenmuZhou committed
257
258
259
    output_names = predictor.get_output_names()
    output_tensors = []
    for output_name in output_names:
WenmuZhou's avatar
WenmuZhou committed
260
        output_tensor = predictor.get_output_handle(output_name)
WenmuZhou's avatar
WenmuZhou committed
261
262
263
264
        output_tensors.append(output_tensor)
    return predictor, input_tensor, output_tensors


Jethong's avatar
Jethong committed
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
def draw_e2e_res(dt_boxes, strs, img_path):
    src_im = cv2.imread(img_path)
    for box, str in zip(dt_boxes, strs):
        box = box.astype(np.int32).reshape((-1, 1, 2))
        cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
        cv2.putText(
            src_im,
            str,
            org=(int(box[0, 0, 0]), int(box[0, 0, 1])),
            fontFace=cv2.FONT_HERSHEY_COMPLEX,
            fontScale=0.7,
            color=(0, 255, 0),
            thickness=1)
    return src_im


LDOUBLEV's avatar
LDOUBLEV committed
281
def draw_text_det_res(dt_boxes, img_path):
LDOUBLEV's avatar
LDOUBLEV committed
282
283
284
285
    src_im = cv2.imread(img_path)
    for box in dt_boxes:
        box = np.array(box).astype(np.int32).reshape(-1, 2)
        cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
LDOUBLEV's avatar
LDOUBLEV committed
286
    return src_im
LDOUBLEV's avatar
LDOUBLEV committed
287
288


LDOUBLEV's avatar
LDOUBLEV committed
289
290
def resize_img(img, input_size=600):
    """
LDOUBLEV's avatar
LDOUBLEV committed
291
    resize img and limit the longest side of the image to input_size
LDOUBLEV's avatar
LDOUBLEV committed
292
293
294
295
296
    """
    img = np.array(img)
    im_shape = img.shape
    im_size_max = np.max(im_shape[0:2])
    im_scale = float(input_size) / float(im_size_max)
WenmuZhou's avatar
WenmuZhou committed
297
298
    img = cv2.resize(img, None, None, fx=im_scale, fy=im_scale)
    return img
LDOUBLEV's avatar
LDOUBLEV committed
299
300


WenmuZhou's avatar
WenmuZhou committed
301
302
303
304
305
def draw_ocr(image,
             boxes,
             txts=None,
             scores=None,
             drop_score=0.5,
LDOUBLEV's avatar
LDOUBLEV committed
306
             font_path="./doc/fonts/simfang.ttf"):
307
308
309
    """
    Visualize the results of OCR detection and recognition
    args:
LDOUBLEV's avatar
LDOUBLEV committed
310
        image(Image|array): RGB image
311
312
313
314
        boxes(list): boxes with shape(N, 4, 2)
        txts(list): the texts
        scores(list): txxs corresponding scores
        drop_score(float): only scores greater than drop_threshold will be visualized
WenmuZhou's avatar
WenmuZhou committed
315
        font_path: the path of font which is used to draw text
316
317
318
    return(array):
        the visualized img
    """
LDOUBLEV's avatar
LDOUBLEV committed
319
320
    if scores is None:
        scores = [1] * len(boxes)
WenmuZhou's avatar
WenmuZhou committed
321
322
323
324
    box_num = len(boxes)
    for i in range(box_num):
        if scores is not None and (scores[i] < drop_score or
                                   math.isnan(scores[i])):
LDOUBLEV's avatar
LDOUBLEV committed
325
            continue
WenmuZhou's avatar
WenmuZhou committed
326
        box = np.reshape(np.array(boxes[i]), [-1, 1, 2]).astype(np.int64)
LDOUBLEV's avatar
LDOUBLEV committed
327
        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)
WenmuZhou's avatar
WenmuZhou committed
328
    if txts is not None:
LDOUBLEV's avatar
LDOUBLEV committed
329
        img = np.array(resize_img(image, input_size=600))
330
        txt_img = text_visual(
WenmuZhou's avatar
WenmuZhou committed
331
332
333
334
335
336
            txts,
            scores,
            img_h=img.shape[0],
            img_w=600,
            threshold=drop_score,
            font_path=font_path)
337
        img = np.concatenate([np.array(img), np.array(txt_img)], axis=1)
LDOUBLEV's avatar
LDOUBLEV committed
338
339
        return img
    return image
340
341


WenmuZhou's avatar
WenmuZhou committed
342
343
344
345
346
347
def draw_ocr_box_txt(image,
                     boxes,
                     txts,
                     scores=None,
                     drop_score=0.5,
                     font_path="./doc/simfang.ttf"):
348
349
350
    h, w = image.height, image.width
    img_left = image.copy()
    img_right = Image.new('RGB', (w, h), (255, 255, 255))
351
352

    import random
LDOUBLEV's avatar
LDOUBLEV committed
353

354
355
356
    random.seed(0)
    draw_left = ImageDraw.Draw(img_left)
    draw_right = ImageDraw.Draw(img_right)
WenmuZhou's avatar
WenmuZhou committed
357
358
359
    for idx, (box, txt) in enumerate(zip(boxes, txts)):
        if scores is not None and scores[idx] < drop_score:
            continue
tink2123's avatar
tink2123 committed
360
361
        color = (random.randint(0, 255), random.randint(0, 255),
                 random.randint(0, 255))
362
        draw_left.polygon(box, fill=color)
tink2123's avatar
tink2123 committed
363
364
365
366
367
368
369
370
371
372
        draw_right.polygon(
            [
                box[0][0], box[0][1], box[1][0], box[1][1], box[2][0],
                box[2][1], box[3][0], box[3][1]
            ],
            outline=color)
        box_height = math.sqrt((box[0][0] - box[3][0])**2 + (box[0][1] - box[3][
            1])**2)
        box_width = math.sqrt((box[0][0] - box[1][0])**2 + (box[0][1] - box[1][
            1])**2)
373
374
        if box_height > 2 * box_width:
            font_size = max(int(box_width * 0.9), 10)
WenmuZhou's avatar
WenmuZhou committed
375
            font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
376
377
378
            cur_y = box[0][1]
            for c in txt:
                char_size = font.getsize(c)
tink2123's avatar
tink2123 committed
379
380
                draw_right.text(
                    (box[0][0] + 3, cur_y), c, fill=(0, 0, 0), font=font)
381
382
383
                cur_y += char_size[1]
        else:
            font_size = max(int(box_height * 0.8), 10)
WenmuZhou's avatar
WenmuZhou committed
384
            font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
tink2123's avatar
tink2123 committed
385
386
            draw_right.text(
                [box[0][0], box[0][1]], txt, fill=(0, 0, 0), font=font)
387
388
389
390
    img_left = Image.blend(image, img_left, 0.5)
    img_show = Image.new('RGB', (w * 2, h), (255, 255, 255))
    img_show.paste(img_left, (0, 0, w, h))
    img_show.paste(img_right, (w, 0, w * 2, h))
391
392
393
    return np.array(img_show)


394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
def str_count(s):
    """
    Count the number of Chinese characters,
    a single English character and a single number
    equal to half the length of Chinese characters.
    args:
        s(string): the input of string
    return(int):
        the number of Chinese characters
    """
    import string
    count_zh = count_pu = 0
    s_len = len(s)
    en_dg_count = 0
    for c in s:
        if c in string.ascii_letters or c.isdigit() or c.isspace():
            en_dg_count += 1
        elif c.isalpha():
            count_zh += 1
        else:
            count_pu += 1
    return s_len - math.ceil(en_dg_count / 2)


WenmuZhou's avatar
WenmuZhou committed
418
419
420
421
422
423
def text_visual(texts,
                scores,
                img_h=400,
                img_w=600,
                threshold=0.,
                font_path="./doc/simfang.ttf"):
424
425
426
427
428
429
430
    """
    create new blank img and draw txt on it
    args:
        texts(list): the text will be draw
        scores(list|None): corresponding score of each txt
        img_h(int): the height of blank img
        img_w(int): the width of blank img
WenmuZhou's avatar
WenmuZhou committed
431
        font_path: the path of font which is used to draw text
432
433
434
435
436
437
438
439
440
    return(array):
    """
    if scores is not None:
        assert len(texts) == len(
            scores), "The number of txts and corresponding scores must match"

    def create_blank_img():
        blank_img = np.ones(shape=[img_h, img_w], dtype=np.int8) * 255
        blank_img[:, img_w - 1:] = 0
LDOUBLEV's avatar
LDOUBLEV committed
441
442
        blank_img = Image.fromarray(blank_img).convert("RGB")
        draw_txt = ImageDraw.Draw(blank_img)
443
        return blank_img, draw_txt
LDOUBLEV's avatar
LDOUBLEV committed
444

445
446
447
448
    blank_img, draw_txt = create_blank_img()

    font_size = 20
    txt_color = (0, 0, 0)
WenmuZhou's avatar
WenmuZhou committed
449
    font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
450
451
452

    gap = font_size + 5
    txt_img_list = []
LDOUBLEV's avatar
LDOUBLEV committed
453
    count, index = 1, 0
454
455
    for idx, txt in enumerate(texts):
        index += 1
LDOUBLEV's avatar
LDOUBLEV committed
456
        if scores[idx] < threshold or math.isnan(scores[idx]):
457
458
459
460
461
462
463
464
465
466
467
            index -= 1
            continue
        first_line = True
        while str_count(txt) >= img_w // font_size - 4:
            tmp = txt
            txt = tmp[:img_w // font_size - 4]
            if first_line:
                new_txt = str(index) + ': ' + txt
                first_line = False
            else:
                new_txt = '    ' + txt
LDOUBLEV's avatar
LDOUBLEV committed
468
            draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
469
470
471
472
473
            txt = tmp[img_w // font_size - 4:]
            if count >= img_h // gap - 1:
                txt_img_list.append(np.array(blank_img))
                blank_img, draw_txt = create_blank_img()
                count = 0
LDOUBLEV's avatar
LDOUBLEV committed
474
            count += 1
475
476
477
        if first_line:
            new_txt = str(index) + ': ' + txt + '   ' + '%.3f' % (scores[idx])
        else:
LDOUBLEV's avatar
LDOUBLEV committed
478
            new_txt = "  " + txt + "  " + '%.3f' % (scores[idx])
LDOUBLEV's avatar
LDOUBLEV committed
479
        draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
480
        # whether add new blank img or not
LDOUBLEV's avatar
LDOUBLEV committed
481
        if count >= img_h // gap - 1 and idx + 1 < len(texts):
482
483
484
            txt_img_list.append(np.array(blank_img))
            blank_img, draw_txt = create_blank_img()
            count = 0
LDOUBLEV's avatar
LDOUBLEV committed
485
        count += 1
486
487
488
489
490
491
    txt_img_list.append(np.array(blank_img))
    if len(txt_img_list) == 1:
        blank_img = np.array(txt_img_list[0])
    else:
        blank_img = np.concatenate(txt_img_list, axis=1)
    return np.array(blank_img)
LDOUBLEV's avatar
LDOUBLEV committed
492
493


dyning's avatar
dyning committed
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
def base64_to_cv2(b64str):
    import base64
    data = base64.b64decode(b64str.encode('utf8'))
    data = np.fromstring(data, np.uint8)
    data = cv2.imdecode(data, cv2.IMREAD_COLOR)
    return data


def draw_boxes(image, boxes, scores=None, drop_score=0.5):
    if scores is None:
        scores = [1] * len(boxes)
    for (box, score) in zip(boxes, scores):
        if score < drop_score:
            continue
        box = np.reshape(np.array(box), [-1, 1, 2]).astype(np.int64)
        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)
    return image


LDOUBLEV's avatar
LDOUBLEV committed
513
if __name__ == '__main__':
LDOUBLEV's avatar
LDOUBLEV committed
514
    pass