准备图像数据(图片特征) 加载模型vgg16,去除最后一层 用该模型提取图片特征(1×4096的矩阵) 将提取的特征保存到名为 “ features.pkl ” 的文件中…
- User
准备图像数据(图片特征)
加载模型vgg16,去除最后一层
用该模型提取图片特征(1×4096的矩阵)
将提取的特征保存到名为 “ features.pkl ” 的文件中
import numpy as np from keras.models import Model from keras.applications.vgg16 import VGG16,preprocess_input from keras.preprocessing.image import load_img,img_to_array from os import listdir from pickle import dump
用vgg16提取图片特征 ,从特定目录
def extract_features(directory): model = VGG16() # 加载模型VGG16 model.layers.pop() # 去掉最后一层 model = Model(inputs=model.inputs, outputs=model.layers[-1].output) model.summary() # 输出summarize
features = {} # 从图片中提取特征,先构建一个空字典 img_namelist = listdir(directory) # 创建图像名列表 for img_name in img_namelist: name = directory +'/'+ img_name # 创建文件名 # 加载图片 img = load_img(name,target_size=(224, 224)) img = img_to_array(img) # 转换为矩阵 img = np.expand_dims(img,axis=0) # 矩阵增维 img = preprocess_input(img) # 预处理:均值化 feature = model.predict(img,verbose=0) # 提取特征并保存 features[img_name[:-4]] = feature # 将向量写入字典 print('>正在处理:',img_name) return featuresdirectory = 'D:/computer Design/look-at-the-picture-and-talk/Flicker8k_Dataset' features = extract_features(directory) dump(features, open('features.pkl', 'wb')) # 保存文件 print('图片特征提取完成,文件已保存!')
- Assistant
加载特征文件并进行预处理
加载图像标题和ID文件
清理文本数据,如将标题转换为小写,删除标点符号和单词长度小于1的单词等
将每个图片的5个标题作为样本,构建字典,以图像ID为key,文本描述为value
将字典保存到名为 “ descriptions.pkl ” 的文件中
from pickle import load,dump from os import listdir from keras.preprocessing.text import Tokenizer
加载已提取特征
def load_features(filename): all_features = load(open(filename, 'rb')) # 仅选择以 .jpg 结尾的文件作为特征 features = {k: all_features[k] for k in listdir('Flicker8k_Dataset') if k+'.jpg' in all_features} return features
加载描述文件
def load_descriptions(filename): file = open(filename, 'r') text = file.read() file.close() descriptions = {} for line in text.split('\n'): tokens = line.split() if len(line) < 2: continue image_id, image_desc = tokens[0], tokens[1:] image_id = image_id.split('.')[0] image_desc = ' '.join(image_desc) if image_id not in descriptions: descriptions[image_id] = [] descriptions[image_id].append(image_desc) return descriptions
清理文本
def clean_descriptions(descriptions): # 转换单词为小写 descriptions = {k:[i.lower() for i in descriptions[k]] for k in descriptions} # 删除单词中的标点符号 table = str.maketrans('','',string.punctuation) for key, desc_list in descriptions.items(): for i in range(len(desc_list)): desc = desc_list[i] desc = desc.split() desc = [word.translate(table) for word in desc] desc = [word for word in desc if len(word)>1] desc = [word for word in desc if word.isalpha()] desc_list[i] = ' '.join(desc) return descriptions
转换字典到文本——用于训练模型
def to_lines(descriptions): all_desc = [] for key in descriptions.keys(): [all_desc.append(d) for d in descriptions[key]] # 存储所有句子 return all_desc
将单词列表转换为token
def create_tokenizer(descriptions): lines = to_lines(descriptions) tokenizer = Tokenizer() tokenizer.fit_on_texts(lines) return tokenizer
保存字典到文件
def save_descriptions(descriptions, filename): lines = to_lines(descriptions) data = '\n'.join(lines) file = open(filename, 'w') file.write(data) file.close()
filename = 'D:/computer Design/look-at-the-picture-and-talk/Flickr8k.token.txt' descriptions = load_descriptions(filename)
清理文本数据
clean_descriptions(descriptions)
将每个图片的5个标题作为一个样本,构建字典,以图像ID为key,文本描述为value
save_descriptions(descriptions, 'descriptions.txt') # 保存数据到文本,用于训练模型
为描述数据生成tokenizer
tokenizer = create_tokenizer(descriptions) dump(tokenizer, open('tokenizer.pkl', 'wb')) # 保存 tokenizer print('数据处理和清理完成')