-
Notifications
You must be signed in to change notification settings - Fork 21
/
Copy pathtest.py
339 lines (298 loc) · 14.5 KB
/
test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
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
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
#!/usr/bin/env python
# -*- coding: utf-8 -*-
from aip import AipOcr
from demo import apiutil
import pylab
import imageio
import skimage
import httplib
from skimage import io,transform,exposure
import numpy as np
import Levenshtein
from pydub import AudioSegment
import collections
import sys
import os
import wave
import argparse
import time
from time import sleep
parser = argparse.ArgumentParser()
#parser.add_argument('--root_dir', type=str,default='/Users/comjia/Downloads/code/pytorch_seq2seq/pytorch_seq2seq',help='输入aishell数据所在的主目录path')
parser.add_argument("--start_df", type=float, default=-0.2,help="字幕开头需要滞后的时间（单位:s),负数表示需要字幕提前出现")
parser.add_argument("--end_df", type=float, default=0.2,help="字幕结尾需要滞后的时间（单位:s),负数表示需要字幕提前结束")
parser.add_argument("--video_name", type=str, default=u'/mnt/steven/data/movie/rmdmy/rmdmy_ep03.mp4',help="视频path")
parser.add_argument("--wav_name", type=str, default=u'/mnt/steven/data/movie/rmdmy/rmdmy_ep03.wav',help="音频path")
parser.add_argument("--ocr_source", type=str, default="tecent",help="baidu or tecent")
parser.add_argument("--movie_name", type=str, default="rmdmy",help="电视剧名称")
parser.add_argument("--srt_and_wav", type=int,default=1,help="是否srt 和 wav同时处理")
parser.add_argument("--account_index", type=int, default=0,help="有三组账号，分别是0，1，2")
opt = parser.parse_args()
print opt
# imageio.plugins.ffmpeg.download()
def contain_chinese_ratio(check_str):
num = 0
for ch in check_str:
if u'\u4e00' <= ch <= u'\u9fff':
num = num+1
return num * 1.0 / len(check_str)
def img_to_str(client,image_path,avg_char_width,flag = 1,retry_times= 5,last_result=""):
with open(image_path, 'rb') as fp:
image = fp.read()
if opt.ocr_source == "baidu":
result = client.basicGeneral(image)
if 'words_result' in result:
if len(result['words_result']) > 2:
return "",None,avg_char_width,result
else:
return '\n'.join([w['words'] for w in result['words_result']]),None,avg_char_width,result
elif opt.ocr_source == "tecent":
if opt.account_index == 0:
app_id = '1106978111'
app_key = '9hUBH27QnbtCWZ2x'
elif opt.account_index == 1:
app_id = '2107593814'
app_key = 'bbUG12n1tnJ9FT8Q'
elif opt.account_index == 2:
app_id = '2107564974'
app_key = 'wRuva6dRRQKD5fCd'
else:
None
while True:
try:
client = apiutil.AiPlat(app_id, app_key)
rsp = client.getOcrGeneralocr(image)
if rsp['ret'] != 0:
return "", None, avg_char_width, rsp
break
except httplib.BadStatusLine :
if retry_times > 0:
retry_times = retry_times - 1
print 'it retry for times:%d' %(retry_times)
sleep(1)
continue
else:
msg = 'it retry 5 times. But it does not run successfully! Please check it.'
raise Exception(msg)
if rsp['ret'] == 0 :
if len(rsp['data']['item_list']) == 1:
#i["itemcoord"][0]["y"],i["itemcoord"][0]["height"]
#找出list中字数最多的那个case,索引放到max_word_index，返回该索引下的字符串和对应的概率
# max_word_list = [len(item["itemstring"]) for item in rsp['data']['item_list'] ]
# max_word_index = max_word_list.index(max(max_word_list))
# fit_sentence = rsp['data']['item_list'][max_word_index]
fit_sentence = rsp['data']['item_list'][0]
print fit_sentence['itemstring'].encode('utf-8')
#如果包含的中文字符少，那么不合格
char_ratio = contain_chinese_ratio(fit_sentence["itemstring"])
if char_ratio <= 0.5:
return "", None, avg_char_width,rsp
# 如果文字宽度不满足要求，也不合格
# 如果宽度符合要求的，那么累积求平均
# if avg_char_width == -1:
# avg_char_width = fit_sentence["itemcoord"][0]["width"] *1.0/len(fit_sentence["itemstring"])
# else:
# if abs( avg_char_width - fit_sentence["itemcoord"][0]["width"] *1.0/len(fit_sentence["itemstring"])) < 3:
#
# avg_char_width += fit_sentence["itemcoord"][0]["width"] *1.0/len(fit_sentence["itemstring"])
# avg_char_width = avg_char_width / 2
# else:
# return "", None, avg_char_width
#返回的文字，需要去除特殊字符
#返回的参数是： 文字，概率，每个字宽
return "".join([cha for cha in fit_sentence["itemstring"] if cha.isalnum()]) \
,[prob["confidence"] for prob in fit_sentence["words"]] \
,avg_char_width,rsp
elif len(rsp['data']['item_list']) > 1:
max_ratio_index = 0
max_ratio = 0
if last_result != "":
for index,i in enumerate(rsp['data']['item_list']):
print i['itemstring'].encode('utf-8')
ratio = Levenshtein.ratio(last_result, i['itemstring'])
if ratio > max_ratio:
max_ratio_index = index
fit_sentence = rsp['data']['item_list'][max_ratio_index]
char_ratio = contain_chinese_ratio(fit_sentence["itemstring"])
if char_ratio <= 0.5:
return "", None, avg_char_width,rsp
else:
return "".join([cha for cha in fit_sentence["itemstring"] if cha.isalnum()]) \
,[prob["confidence"] for prob in fit_sentence["words"]] \
,avg_char_width,rsp
else:
return "",None,avg_char_width,rsp
#对于 一段文字在多帧展示出来，ocr识别了多帧结果，使用list保存下来，使用规则根据ocr的prob来得到最准确的该段文字
#规则如下：
#使用 两个map的list来分别保存字出现的次数和字出现的概率之和，返回字个数多的；当字的个数相同时，返回字打分高的
def same_rule(same_list):
max_len = 0
#
for item in same_list:
if len(item[0]) > max_len:
max_len = len(item[0])
char_map_list = []
char_map_score = []
for i in range(max_len):
char_map_list.append(dict())
char_map_score.append(dict())
for index in range(max_len):
for i in range(len(same_list)):
if index < len(same_list[i][0]):
char = same_list[i][0][index]
score = same_list[i][1][index]
else:
char = ""
score = 1
if char not in char_map_list[index]:
char_map_list[index][char] = 1
else:
char_map_list[index][char] += 1
if char not in char_map_score[index]:
char_map_score[index][char] = score
else:
char_map_score[index][char] += score
#返回个数多的；当个数相同时，返回打分高的
result = []
for i in range(max_len):
result.append(sorted(char_map_list[i].items(),key=lambda item:item[1],reverse=True)[0][0])
return "".join(result)
#根据文字切割音频段
def cut_video_based_filelist(srt_filename,wav_name):
np_vec = np.load(srt_filename)
time_stamp_list = np_vec["filelist"]
path = wav_name
format_type = path.split(".")[-1]
if format_type == "wav":
wav_audio = AudioSegment.from_file(path, format="wav").set_channels(1)
elif format_type == "mp3":
wav_audio = AudioSegment.from_file(path, format="mp3").set_channels(1)
elif format_type == "m4a":
wav_audio = AudioSegment.from_file(path, format="mp4").set_channels(1)
# audio = wav_audio.raw_data
# sample_rate = wav_audio.frame_rate
video_name = opt.video_name.split("/")[-1].split(".")[-2]
audio_dir_name = opt.ocr_source +"_"+ video_name
os.mkdir(video_name)
start_df = opt.start_df
end_df = opt.end_df
with open(video_name+".txt","w") as f:
for index,time_stamp in enumerate(time_stamp_list):
if abs(float(time_stamp[1]) - float(time_stamp[0]) ) < 0.0001 :
continue
# starttime_stamp = int(((float(time_stamp[0]) - srt_df) * sample_rate * 2))
# endtime_stamp = int(((float(time_stamp[1]) + srt_df) * sample_rate * 2))
# audio_seg = audio[starttime_stamp:endtime_stamp]
#假如当前的字幕结束时间比下一帧字幕的开始时间还要迟，就以下一帧字幕的开始时间作为当前帧字幕的结束时间
if index < len(time_stamp_list) - 1 and (float(time_stamp[1]) + end_df) * 1000 > (float(time_stamp_list[index+1][0])+start_df) * 1000:
end_stamp = (float(time_stamp_list[index+1][0])+start_df) * 1000
else:
end_stamp = (float(time_stamp[1]) + end_df) * 1000
audio_seg = wav_audio[(float(time_stamp[0]) + start_df) * 1000:end_stamp]
save_name = '%s/%0004d.wav' % (video_name,index)
text = "".join([cha for cha in (time_stamp[2]) if cha.isalnum()])
f.write(",".join(["/movie/wav/%s/%s/%04d.wav"%(opt.movie_name,video_name,index),text.encode("utf-8")])+"\n")
audio_seg.export(save_name,format ="wav")
#从视频中提取文字
def pull_srt_from_video(video_name,save_srt_name):
if opt.ocr_source == "baidu":
APP_ID = '11531274'
API_KEY = 'nl59T9O2lmZ7iAD2wttS457F'
SECRET_KEY = 'U0VztUf0QKwjfTxzxIcG1CWf9qz9Sobf'
client = AipOcr(APP_ID, API_KEY, SECRET_KEY)
elif opt.ocr_source == "tecent":
app_id = '1106978111'
app_key = '9hUBH27QnbtCWZ2x'
client = apiutil.AiPlat(app_id, app_key)
vid = imageio.get_reader(video_name, 'ffmpeg')
all_frames = vid.get_length()
filelist = [[0,0,u""]]#格式为：[starttime , endtime, data]
interval_frame = 5
framerate = vid.get_meta_data()['fps']
same_list = []
avg_char_width = -1
if opt.movie_name == "rmdmy":
start_frame = (1 * 60 + 30 ) * 25
end_frame = (43 * 60 + 30 ) * 25
elif opt.movie_name == "wdqbs":
start_frame = (1 * 60 + 30 ) * 25
end_frame = all_frames - start_frame
elif opt.movie_name == "bly":
start_frame = (2 * 60 + 30 ) * 25
end_frame = all_frames - ((3 * 60 + 40 ) * 25)
elif opt.movie_name == "nrb":
start_frame = (0 * 60 + 5 ) * 25
end_frame = all_frames
print "all_frames = %d,end_frame = %d" %(all_frames,end_frame)
last_result = ""
try:
for num in range(all_frames):
if num < start_frame or num > end_frame:
continue
im = vid.get_data(num)
if num % interval_frame != 0: #每10帧是40ms，帧率是25hz
continue
print num
image = im#skimage.img_as_float(im).astype(np.float64)
# if flag == 0:
# imageio.imsave("abcd.jpg", image[image.shape[0] * 2 /3:,:])
# words_tmp, porb_tmp, item = img_to_str(client, "abcd.jpg", avg_char_width,flag=flag)
# if words_tmp=="":
# continue
# y = item[0]["y"]
# height = item[0]["height"]
# if y > 0:
# flag = 1
if opt.movie_name == "rmdmy":
tmp = image[(image.shape[0] * 2 / 3 + 100):(image.shape[0] * 2 / 3 + 200), 200:1000]
elif opt.movie_name == "wdqbs":
tmp = image[(image.shape[0] * 2 /3 + 180):(image.shape[0] * 2 /3 + 320),int(image.shape[1] * 0.15) :int(image.shape[1] * 0.78)]
elif opt.movie_name == "bly":
tmp = image[(image.shape[0] * 2 / 3 + 180):(image.shape[0] * 2 / 3 + 320),
int(image.shape[1] * 0.15):int(image.shape[1] * 0.78)]
elif opt.movie_name == "nrb":
tmp = image[(image.shape[0] * 2 / 3 + 120):(image.shape[0] * 2 / 3 + 320),
int(image.shape[1] * 0.17):int(image.shape[1] * 0.88)]
save_file_name = "%s_%d.jpg"%(save_srt_name,num)
imageio.imsave(save_file_name,tmp)
result , porb,avg_char_width, rsp = img_to_str(None,save_file_name,avg_char_width,last_result)
if rsp['ret'] != 0 or (rsp['ret'] == 0 and result == ""):
gam2 = exposure.adjust_gamma(tmp, 0.5)
imageio.imsave(save_file_name, gam2)
result , porb,avg_char_width, rsp = img_to_str(None,save_file_name,avg_char_width,last_result)
last_result = result
os.remove(save_file_name)
print result.encode("utf-8")
if result != "" :
if len(same_list) > 0:
ratio = Levenshtein.ratio(result, filelist[-1][-1])
if ratio < 0.5:
filelist[-1][2] = same_rule(same_list)
same_list = [(result, porb)]
filelist.append(
[num * 1.0 / framerate, num * 1.0 / framerate, result])
else:
same_list.append((result,porb))
filelist[-1][1] = num * 1.0 / framerate
else:
same_list = [(result, porb)]
filelist.append(
[num * 1.0 / framerate, num * 1.0 / framerate, result]) # 格式为：[starttime , endtime, data]
#for reply system busy
# elif result == "" and porb == -1:
# None
else:
if len(same_list) > 0:
filelist[-1][2] = same_rule(same_list)
same_list = []
except IOError:
print('可能是ioerror ')
finally:
np.savez(save_srt_name,
filelist=filelist
)
if __name__ == '__main__':
save_srt_name = "%s_%s_movie_srt.npz" % (opt.video_name.split("/")[-1].split(".")[-2],opt.ocr_source)
if opt.srt_and_wav == 1:
pull_srt_from_video(video_name = opt.video_name,save_srt_name = save_srt_name )
cut_video_based_filelist(srt_filename=save_srt_name,wav_name = opt.wav_name)