-
Notifications
You must be signed in to change notification settings - Fork 0
/
extractify.py
137 lines (118 loc) · 4.39 KB
/
extractify.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
from transformers import BertTokenizer
import tensorflow as tf
import numpy as np
import json
import gdown
import os
#bert_classifier Class
class bert_classifier:
def __init__(self):
path='./static/MODELS/bert_classifier'
url='https://drive.google.com/drive/folders/1u8pPnb2qPTt67Yf3v8enC2pF2HpyWaEc'
if not os.path.isdir(path):
gdown.download_folder(url,output=path, quiet=False)
self.tokenizer = BertTokenizer.from_pretrained('bert-base-cased')
self.model = tf.keras.models.load_model(path)
def preprocessing(self, input_data):
tokens = self.tokenizer.encode_plus(input_data, max_length=50,
truncation=True, padding='max_length',
add_special_tokens=True, return_token_type_ids=False,
return_tensors='tf')
# tokenizer returns int32 tensors, we need to return float64, so we use tf.cast
in_tensor = {'input_ids': tf.cast(tokens['input_ids'], tf.float64),
'attention_mask': tf.cast(tokens['attention_mask'], tf.float64)}
return in_tensor
def predict(self, input_data):
return self.model.predict(input_data)[0]
def postprocessing(self, input_data):
bins={1:'key',2:'value',0:'other'}
return bins[np.argmax(input_data)],str(np.max(input_data))
def compute_prediction(self, input_data):
try:
input_data = self.preprocessing(input_data)
prediction = self.predict(input_data)
prediction = self.postprocessing(prediction)
except Exception as e:
return {"status": "Error", "message": str(e)}
return prediction
#linking Classifier Class
class linking_classifier:
def __init__(self):
path='./static/MODELS/linking'
url='https://drive.google.com/drive/folders/1vDUBSS6cXyZdQTwtcXhg8BXBGGR_s7FT'
if not os.path.isdir(path):
gdown.download_folder(url,output=path, quiet=False)
self.model = tf.keras.models.load_model(path)
def ext(self,box):
width=box[2]-box[0]
height=box[3]-box[1]
c_x=(box[0]+box[2])/2
c_y=(box[1]+box[3])/2
return c_x,c_y,width,height
def preprocessing(self, input_data1,input_data2):
b1=self.ext(input_data1)
b2=self.ext(input_data2)
dis=(b1[0]-b2[0])**2 + (b1[1]-b2[1])**2
dis=dis**0.5
sin=(b1[1]-b2[1])/dis
sin*= -1 if sin<0 else 1
relative_width = b1[2]/b2[2] if b2[2]>b1[2] else b2[2]/b1[2]
relative_height = b1[3]/b2[3] if b2[3]>b1[3] else b2[3]/b1[3]
return [dis/1000,sin,relative_width,relative_height]
def predict(self, input_data):
return self.model.predict(np.array([input_data]))[0]
def postprocessing(self, input_data):
bins={1:'linked',0:'unlinked'}
return bins[np.argmax(input_data)],str(np.max(input_data))
def compute_prediction(self, input_data1,input_data2):
try:
input_data = self.preprocessing(input_data1,input_data2)
prediction = self.predict(input_data)
prediction = self.postprocessing(prediction)
except Exception as e:
return {"status": "Error", "message": str(e)}
return prediction
#Wrapper Class
class Wrapper:
def __init__(self,classifier,linker):
self.data = None
self.classifier = classifier
self.linker=linker
self.keys=[]
self.values=[]
def preprocessing(self,json):
self.data = None
self.data=json
self.values=[]
self.keys=[]
try:
for i in self.data:
del i['label']
del i ['linking']
except:
pass
def classification(self):
for i in self.data['form']:
i['linking']=[]
i['label']=list(self.classifier.compute_prediction(i['text']))
if i['label'][0]=='key':
self.keys.append(i['id'])
if i['label'][0]=='value':
self.values.append(i['id'])
def linking(self):
for i in self.keys:
for j in self.values:
link = list(self.linker.compute_prediction(self.data['form'][i]['box'],self.data['form'][j]['box']))
if link[0] == 'linked':
self.data['form'][i]['linking'].append([i,j,link[1]])
self.data['form'][j]['linking'].append([i,j,link[1]])
def generate(self,json):
self.preprocessing(json)
self.classification()
self.linking()
return self.data
classifier = bert_classifier()
linker = linking_classifier()
wrapper = Wrapper(classifier,linker)
def generate(json):
return wrapper.generate(json)