国产片侵犯亲女视频播放_亚洲精品二区_在线免费国产视频_欧美精品一区二区三区在线_少妇久久久_在线观看av不卡

腳本之家,腳本語言編程技術及教程分享平臺!
分類導航

Python|VBS|Ruby|Lua|perl|VBA|Golang|PowerShell|Erlang|autoit|Dos|bat|

服務器之家 - 腳本之家 - Python - keras實現基于孿生網絡的圖片相似度計算方式

keras實現基于孿生網絡的圖片相似度計算方式

2020-06-12 10:39黑條紋的白斑馬 Python

這篇文章主要介紹了keras實現基于孿生網絡的圖片相似度計算方式,具有很好的參考價值,希望對大家有所幫助。一起跟隨小編過來看看吧

我就廢話不多說了,大家還是直接看代碼吧!

?
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
import keras
from keras.layers import Input,Dense,Conv2D
from keras.layers import MaxPooling2D,Flatten,Convolution2D
from keras.models import Model
import os
import numpy as np
from PIL import Image
from keras.optimizers import SGD
from scipy import misc
root_path = os.getcwd()
train_names = ['bear','blackswan','bus','camel','car','cows','dance','dog','hike','hoc','kite','lucia','mallerd','pigs','soapbox','stro','surf','swing','train','walking']
test_names = ['boat','dance-jump','drift-turn','elephant','libby']
 
def load_data(seq_names,data_number,seq_len):
#生成圖片對
  print('loading data.....')
  frame_num = 51
  train_data1 = []
  train_data2 = []
  train_lab = []
  count = 0
  while count < data_number:
    count = count + 1
    pos_neg = np.random.randint(0,2)
    if pos_neg==0:
      seed1 = np.random.randint(0,seq_len)
      seed2 = np.random.randint(0,seq_len)
      while seed1 == seed2:
       seed1 = np.random.randint(0,seq_len)
       seed2 = np.random.randint(0,seq_len)
      frame1 = np.random.randint(1,frame_num)
      frame2 = np.random.randint(1,frame_num)
      path1 = os.path.join(root_path,'data','simility_data',seq_names[seed1],str(frame1)+'.jpg')
      path2 = os.path.join(root_path, 'data', 'simility_data', seq_names[seed2], str(frame2) + '.jpg')
      image1 = np.array(misc.imresize(Image.open(path1),[224,224]))
      image2 = np.array(misc.imresize(Image.open(path2),[224,224]))
      train_data1.append(image1)
      train_data2.append(image2)
      train_lab.append(np.array(0))
    else:
     seed = np.random.randint(0,seq_len)
     frame1 = np.random.randint(1, frame_num)
     frame2 = np.random.randint(1, frame_num)
     path1 = os.path.join(root_path, 'data', 'simility_data', seq_names[seed], str(frame1) + '.jpg')
     path2 = os.path.join(root_path, 'data', 'simility_data', seq_names[seed], str(frame2) + '.jpg')
     image1 = np.array(misc.imresize(Image.open(path1),[224,224]))
     image2 = np.array(misc.imresize(Image.open(path2),[224,224]))
     train_data1.append(image1)
     train_data2.append(image2)
     train_lab.append(np.array(1))
  return np.array(train_data1),np.array(train_data2),np.array(train_lab)
 
def vgg_16_base(input_tensor):
  net = Conv2D(64(3,3),activation='relu',padding='same',input_shape=(224,224,3))(input_tensor)
  net = Convolution2D(64,(3,3),activation='relu',padding='same')(net)
  net = MaxPooling2D((2,2),strides=(2,2))(net)
 
  net = Convolution2D(128,(3,3),activation='relu',padding='same')(net)
  net = Convolution2D(128,(3,3),activation='relu',padding='same')(net)
  net= MaxPooling2D((2,2),strides=(2,2))(net)
 
  net = Convolution2D(256,(3,3),activation='relu',padding='same')(net)
  net = Convolution2D(256,(3,3),activation='relu',padding='same')(net)
  net = Convolution2D(256,(3,3),activation='relu',padding='same')(net)
  net = MaxPooling2D((2,2),strides=(2,2))(net)
 
  net = Convolution2D(512,(3,3),activation='relu',padding='same')(net)
  net = Convolution2D(512,(3,3),activation='relu',padding='same')(net)
  net = Convolution2D(512,(3,3),activation='relu',padding='same')(net)
  net = MaxPooling2D((2,2),strides=(2,2))(net)
 
  net = Convolution2D(512,(3,3),activation='relu',padding='same')(net)
  net = Convolution2D(512,(3,3),activation='relu',padding='same')(net)
  net = Convolution2D(512,(3,3),activation='relu',padding='same')(net)
  net = MaxPooling2D((2,2),strides=(2,2))(net)
  net = Flatten()(net)
  return net
 
def siamese(vgg_path=None,siamese_path=None):
  input_tensor = Input(shape=(224,224,3))
  vgg_model = Model(input_tensor,vgg_16_base(input_tensor))
  if vgg_path:
    vgg_model.load_weights(vgg_path)
  input_im1 = Input(shape=(224,224,3))
  input_im2 = Input(shape=(224,224,3))
  out_im1 = vgg_model(input_im1)
  out_im2 = vgg_model(input_im2)
  diff = keras.layers.substract([out_im1,out_im2])
  out = Dense(500,activation='relu')(diff)
  out = Dense(1,activation='sigmoid')(out)
  model = Model([input_im1,input_im2],out)
  if siamese_path:
    model.load_weights(siamese_path)
  return model
 
train = True
if train:
  model = siamese(siamese_path='model/simility/vgg.h5')
  sgd = SGD(lr=1e-6,momentum=0.9,decay=1e-6,nesterov=True)
  model.compile(optimizer=sgd,loss='mse',metrics=['accuracy'])
  tensorboard = keras.callbacks.TensorBoard(histogram_freq=5,log_dir='log/simility',write_grads=True,write_images=True)
  ckpt = keras.callbacks.ModelCheckpoint(os.path.join(root_path,'model','simility','vgg.h5'),
                    verbose=1,period=5)
  train_data1,train_data2,train_lab = load_data(train_names,4000,20)
  model.fit([train_data1,train_data2],train_lab,callbacks=[tensorboard,ckpt],batch_size=64,epochs=50)
else:
  model = siamese(siamese_path='model/simility/vgg.h5')
  test_im1,test_im2,test_labe = load_data(test_names,1000,5)
  TP = 0
  for i in range(1000):
   im1 = np.expand_dims(test_im1[i],axis=0)
   im2 = np.expand_dims(test_im2[i],axis=0)
   lab = test_labe[i]
   pre = model.predict([im1,im2])
   if pre>0.9 and lab==1:
    TP = TP + 1
   if pre<0.9 and lab==0:
    TP = TP + 1
  print(float(TP)/1000)

輸入兩張圖片,標記1為相似,0為不相似。

損失函數用的是簡單的均方誤差,有待改成Siamese的對比損失。

總結:

1.隨機生成了幾組1000對的圖片,測試精度0.7左右,效果一般。

2.問題 1)數據加載沒有用生成器,還得繼續認真看看文檔 2)訓練時劃分驗證集的時候,訓練就會報錯,什么輸入維度的問題,暫時沒找到原因 3)輸入的shape好像必須給出數字,本想用shape= input_tensor.get_shape(),能訓練,不能保存模型,會報(NOT JSON Serializable,Dimension(None))類型錯誤

補充知識: keras 問答匹配孿生網絡文本匹配 RNN 帶有數據

用途:

這篇博客解釋了如何搭建一個簡單的匹配網絡。并且使用了keras的lambda層。在建立網絡之前需要對數據進行預處理。處理過后,文本轉變為id字符序列。將一對question,answer分別編碼可以得到兩個向量,在匹配層中比較兩個向量,計算相似度。

網絡圖示:

keras實現基于孿生網絡的圖片相似度計算方式

數據準備:

數據基于網上的淘寶客服對話數據,我也會放在我的下載頁面中。原數據是對話,我篩選了其中label為1的對話。然后將對話拆解成QA對,q是用戶,a是客服。然后對于每個q,有一個a是匹配的,label為1.再選擇一個a,構成新的樣本,label為0.

超參數:

比較簡單,具體看代碼就可以了。

?
1
2
3
4
5
6
7
8
9
10
11
# dialogue max pair q,a
max_pair = 30000
# top k frequent word ,k
MAX_FEATURES = 450
# fixed q,a length
MAX_SENTENCE_LENGTH = 30
embedding_size = 100
batch_size = 600
# learning rate
lr = 0.01
HIDDEN_LAYER_SIZE = n_hidden_units = 256 # neurons in hidden layer

細節:

導入一些庫

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
# -*- coding: utf-8 -*-
from keras.layers.core import Activation, Dense, Dropout, SpatialDropout1D
from keras.layers.embeddings import Embedding
from keras.layers.recurrent import LSTM
from keras.preprocessing import sequence
from sklearn.model_selection import train_test_split
import collections
import matplotlib.pyplot as plt
import nltk
import numpy as np
import os
import pandas as pd
from alime_data import convert_dialogue_to_pair
from parameter import MAX_SENTENCE_LENGTH,MAX_FEATURES,embedding_size,max_pair,batch_size,HIDDEN_LAYER_SIZE
DATA_DIR = "../data"
NUM_EPOCHS = 2
# Read training data and generate vocabulary
maxlen = 0
num_recs = 0

數據準備,先統計詞頻,然后取出top N個常用詞,然后將句子轉換成 單詞id的序列。把句子中的有效id靠右邊放,將句子左邊補齊padding。然后分成訓練集和測試集

?
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
word_freqs = collections.Counter()
training_data = convert_dialogue_to_pair(max_pair)
num_recs = len([1 for r in training_data.iterrows()])
 
#for line in ftrain:
for line in training_data.iterrows():
  label ,sentence_q = line[1]['label'],line[1]['sentence_q']
  label ,sentence_a = line[1]['label'],line[1]['sentence_a']
  words = nltk.word_tokenize(sentence_q.lower())#.decode("ascii", "ignore")
  if len(words) > maxlen:
    maxlen = len(words)
  for word in words:
    word_freqs[word] += 1
  words = nltk.word_tokenize(sentence_a.lower())#.decode("ascii", "ignore")
  if len(words) > maxlen:
    maxlen = len(words)
  for word in words:
    word_freqs[word] += 1
  #num_recs += 1
## Get some information about our corpus
 
# 1 is UNK, 0 is PAD
# We take MAX_FEATURES-1 featurs to accound for PAD
vocab_size = min(MAX_FEATURES, len(word_freqs)) + 2
word2index = {x[0]: i+2 for i, x in enumerate(word_freqs.most_common(MAX_FEATURES))}
word2index["PAD"] = 0
word2index["UNK"] = 1
index2word = {v:k for k, v in word2index.items()}
# convert sentences to sequences
X_q = np.empty((num_recs, ), dtype=list)
X_a = np.empty((num_recs, ), dtype=list)
y = np.zeros((num_recs, ))
i = 0
def chinese_split(x):
  return x.split(' ')
 
for line in training_data.iterrows():
  label ,sentence_q,sentence_a = line[1]['label'],line[1]['sentence_q'],line[1]['sentence_a']
  #label, sentence = line.strip().split("\t")
  #print(label,sentence)
  #words = nltk.word_tokenize(sentence_q.lower())
  words = chinese_split(sentence_q)
  seqs = []
  for word in words:
    if word in word2index.keys():
      seqs.append(word2index[word])
    else:
      seqs.append(word2index["UNK"])
  X_q[i] = seqs
  #print('add_q')
  #words = nltk.word_tokenize(sentence_a.lower())
  words = chinese_split(sentence_a)
  seqs = []
  for word in words:
    if word in word2index.keys():
      seqs.append(word2index[word])
    else:
      seqs.append(word2index["UNK"])
  X_a[i] = seqs
  y[i] = int(label)
  i += 1
# Pad the sequences (left padded with zeros)
X_a = sequence.pad_sequences(X_a, maxlen=MAX_SENTENCE_LENGTH)
X_q = sequence.pad_sequences(X_q, maxlen=MAX_SENTENCE_LENGTH)
X = []
for i in range(len(X_a)):
  concat = [X_q[i],X_a[i]]
  X.append(concat)
 
# Split input into training and test
Xtrain, Xtest, ytrain, ytest = train_test_split(X, y, test_size=0.2,
                        random_state=42)
#print(Xtrain.shape, Xtest.shape, ytrain.shape, ytest.shape)
Xtrain_Q = [e[0] for e in Xtrain]
Xtrain_A = [e[1] for e in Xtrain]
Xtest_Q = [e[0] for e in Xtest]
Xtest_A = [e[1] for e in Xtest]

最后建立網絡。先定義兩個函數,一個是句子編碼器,另一個是lambda層,計算兩個向量的絕對差。將QA分別用encoder處理得到兩個向量,把兩個向量放入lambda層。最后有了2*hidden size的一層,將這一層接一個dense層,接activation,得到分類概率。

?
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
from keras.layers.wrappers import Bidirectional
from keras.layers import Input,Lambda
from keras.models import Model
 
def encoder(inputs_seqs,rnn_hidden_size,dropout_rate):
  x_embed = Embedding(vocab_size, embedding_size, input_length=MAX_SENTENCE_LENGTH)(inputs_seqs)
  inputs_drop = SpatialDropout1D(0.2)(x_embed)
  encoded_Q = Bidirectional(
    LSTM(rnn_hidden_size, dropout=dropout_rate, recurrent_dropout=dropout_rate, name='RNN'))(inputs_drop)
  return encoded_Q
 
def absolute_difference(vecs):
  a,b =vecs
  #d = a-b
  return abs(a - b)
 
inputs_Q = Input(shape=(MAX_SENTENCE_LENGTH,), name="input")
# x_embed = Embedding(vocab_size, embedding_size, input_length=MAX_SENTENCE_LENGTH)(inputs_Q)
# inputs_drop = SpatialDropout1D(0.2)(x_embed)
# encoded_Q = Bidirectional(LSTM(HIDDEN_LAYER_SIZE, dropout=0.2, recurrent_dropout=0.2,name= 'RNN'))(inputs_drop)
inputs_A = Input(shape=(MAX_SENTENCE_LENGTH,), name="input_a")
# x_embed = Embedding(vocab_size, embedding_size, input_length=MAX_SENTENCE_LENGTH)(inputs_A)
# inputs_drop = SpatialDropout1D(0.2)(x_embed)
# encoded_A = Bidirectional(LSTM(HIDDEN_LAYER_SIZE, dropout=0.2, recurrent_dropout=0.2,name= 'RNN'))(inputs_drop)
encoded_Q = encoder(inputs_Q,HIDDEN_LAYER_SIZE,0.1)
encoded_A = encoder(inputs_A,HIDDEN_LAYER_SIZE,0.1)
 
# import tensorflow as tf
# difference = tf.subtract(encoded_Q, encoded_A)
# difference = tf.abs(difference)
similarity = Lambda(absolute_difference)([encoded_Q, encoded_A])
# x = concatenate([encoded_Q, encoded_A])
#
# matching_x = Dense(128)(x)
# matching_x = Activation("sigmoid")(matching_x)
polar = Dense(1)(similarity)
prop = Activation("sigmoid")(polar)
model = Model(inputs=[inputs_Q,inputs_A], outputs=prop)
model.compile(loss="binary_crossentropy", optimizer="adam",
       metrics=["accuracy"])
training_history = model.fit([Xtrain_Q, Xtrain_A], ytrain, batch_size=batch_size,
               epochs=NUM_EPOCHS,
               validation_data=([Xtest_Q,Xtest_A], ytest))
# plot loss and accuracy
def plot(training_history):
  plt.subplot(211)
  plt.title("Accuracy")
  plt.plot(training_history.history["acc"], color="g", label="Train")
  plt.plot(training_history.history["val_acc"], color="b", label="Validation")
  plt.legend(loc="best")
 
  plt.subplot(212)
  plt.title("Loss")
  plt.plot(training_history.history["loss"], color="g", label="Train")
  plt.plot(training_history.history["val_loss"], color="b", label="Validation")
  plt.legend(loc="best")
  plt.tight_layout()
  plt.show()
 
# evaluate
score, acc = model.evaluate([Xtest_Q,Xtest_A], ytest, batch_size = batch_size)
print("Test score: %.3f, accuracy: %.3f" % (score, acc))
 
for i in range(25):
  idx = np.random.randint(len(Xtest_Q))
  #idx2 = np.random.randint(len(Xtest_A))
  xtest_Q = Xtest_Q[idx].reshape(1,MAX_SENTENCE_LENGTH)
  xtest_A = Xtest_A[idx].reshape(1,MAX_SENTENCE_LENGTH)
  ylabel = ytest[idx]
  ypred = model.predict([xtest_Q,xtest_A])[0][0]
  sent_Q = " ".join([index2word[x] for x in xtest_Q[0].tolist() if x != 0])
  sent_A = " ".join([index2word[x] for x in xtest_A[0].tolist() if x != 0])
  print("%.0f\t%d\t%s\t%s" % (ypred, ylabel, sent_Q,sent_A))

最后是處理數據的函數,寫在另一個文件里。

?
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
import nltk
from parameter import MAX_FEATURES,MAX_SENTENCE_LENGTH
import pandas as pd
from collections import Counter
def get_pair(number, dialogue):
  pairs = []
  for conversation in dialogue:
    utterances = conversation[2:].strip('\n').split('\t')
    # print(utterances)
    # break
 
    for i, utterance in enumerate(utterances):
      if i % 2 != 0: continue
      pairs.append([utterances[i], utterances[i + 1]])
      if len(pairs) >= number:
        return pairs
  return pairs
 
 
def convert_dialogue_to_pair(k):
  dialogue = open('dialogue_alibaba2.txt', encoding='utf-8', mode='r')
  dialogue = dialogue.readlines()
  dialogue = [p for p in dialogue if p.startswith('1')]
  print(len(dialogue))
  pairs = get_pair(k, dialogue)
  # break
  # print(pairs)
  data = []
  for p in pairs:
    data.append([p[0], p[1], 1])
  for i, p in enumerate(pairs):
    data.append([p[0], pairs[(i + 8) % len(pairs)][1], 0])
  df = pd.DataFrame(data, columns=['sentence_q', 'sentence_a', 'label'])
 
  print(len(data))
  return df

以上這篇keras實現基于孿生網絡的圖片相似度計算方式就是小編分享給大家的全部內容了,希望能給大家一個參考,也希望大家多多支持服務器之家。

原文鏈接:https://blog.csdn.net/u013491950/article/details/83997234

延伸 · 閱讀

精彩推薦
主站蜘蛛池模板: 国产精品3区 | 国产精品九九久久99视频 | 欧美日韩精品一区 | 久re在线| 一区亚洲 | 一区二区三区在线不卡 | 日韩一区二区不卡 | 一卡二卡久久 | 不卡av一区二区三区 | 久久精品中文字幕 | 一区自拍 | 国产成人精品久久 | 国产综合亚洲精品一区二 | 在线影院av | 久久久久av| 国产高清在线观看 | 秋霞特色aa大片 | 在线观看91视频 | av7777 | 日本一区二区不卡在线观看 | 日韩欧美国产一区二区 | 精品午夜久久 | 九九99九九| 欧美一级二级视频 | 在线91| 国产96在线视频 | 欧美日韩中文字幕在线 | 亚洲欧美激情精品一区二区 | 毛片国产| 毛片免费观看视频 | 亚洲精品一区中文字幕乱码 | 亚洲精品一区二区三区在线 | 天堂av中文字幕 | 亚洲国产一级毛片 | 四虎久久精品 | 精品免费国产一区二区三区四区 | 日韩精品在线观看视频 | www伊人 | 成年人激情视频 | 日韩精品无码一区二区三区 | 久久免费看少妇a高潮一片黄特 |