【机器学习应用】入门实验：基于vgg网络和Keras框架实现“以图搜图”

# -*- coding: utf-8 -*-
import numpy as np
from numpy import linalg as LA
 
from keras.applications.vgg16 import VGG16
from keras.preprocessing import image
from keras.applications.vgg16 import preprocess_input as preprocess_input_vgg
class VGGNet:
    def __init__(self):
        self.input_shape = (224, 224, 3)
        self.weight = 'imagenet'
        self.pooling = 'max'
        self.model_vgg = VGG16(weights = self.weight, input_shape = (self.input_shape[0], self.input_shape[1], self.input_shape[2]), pooling = self.pooling, include_top = False)
        self.model_vgg.predict(np.zeros((1, 224, 224 , 3)))
 
    #提取vgg16最后一层卷积特征
    def vgg_extract_feat(self, img_path):
        img = image.load_img(img_path, target_size=(self.input_shape[0], self.input_shape[1]))
        img = image.img_to_array(img)
        img = np.expand_dims(img, axis=0)
        img = preprocess_input_vgg(img)
        feat = self.model_vgg.predict(img)
        # print(feat.shape)
        norm_feat = feat[0]/LA.norm(feat[0])
        return norm_feat

# -*- coding: utf-8 -*-
import os
import h5py
import numpy as np
import argparse
from vgg import VGGNet
 
def get_imlist(path):
    return [os.path.join(path, f) for f in os.listdir(path) if f.endswith('.jpg')]
 
if __name__ == "__main__":
    database = r'D:\pythonProject5\flower_roses'
    index = 'vgg_featureCNN.h5'
    img_list = get_imlist(database)
 
    print("         feature extraction starts")
 
    feats = []
    names = []
 
    model = VGGNet()
    for i, img_path in enumerate(img_list):
        norm_feat = model.vgg_extract_feat(img_path)  # 修改此处改变提取特征的网络
        img_name = os.path.split(img_path)[1]
        feats.append(norm_feat)
        names.append(img_name)
        print("extracting feature from image No. %d , %d images in total" % ((i + 1), len(img_list)))
 
    feats = np.array(feats)
 
    output = index
    print("      writing feature extraction results ...")
 
    h5f = h5py.File(output, 'w')
    h5f.create_dataset('dataset_1', data=feats)
    # h5f.create_dataset('dataset_2', data = names)
    h5f.create_dataset('dataset_2', data=np.string_(names))
    h5f.close()

# -*- coding: utf-8 -*-
from vgg import VGGNet
import numpy as np
import h5py
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import argparse
 
query = r'D:\pythonProject5\rose\red_rose.jpg'
index = 'vgg_featureCNN.h5'
result = r'D:\pythonProject5\flower_roses'
# read in indexed images' feature vectors and corresponding image names
h5f = h5py.File(index, 'r')
# feats = h5f['dataset_1'][:]
feats = h5f['dataset_1'][:]
print(feats)
imgNames = h5f['dataset_2'][:]
print(imgNames)
h5f.close()
print("               searching starts")
queryImg = mpimg.imread(query)
plt.title("Query Image")
plt.imshow(queryImg)
plt.show()
 
# init VGGNet16 model
model = VGGNet()
# extract query image's feature, compute simlarity score and sort
queryVec = model.vgg_extract_feat(query)  # 修改此处改变提取特征的网络
print(queryVec.shape)
print(feats.shape)
scores = np.dot(queryVec, feats.T)
rank_ID = np.argsort(scores)[::-1]
rank_score = scores[rank_ID]
# print (rank_ID)
print(rank_score)
# number of top retrieved images to show
maxres = 6  # 检索出6张相似度最高的图片
imlist = []
for i, index in enumerate(rank_ID[0:maxres]):
    imlist.append(imgNames[index])
    print(type(imgNames[index]))
    print("image names: " + str(imgNames[index]) + " scores: %f" % rank_score[i])
print("top %d images in order are: " % maxres, imlist)
# show top #maxres retrieved result one by one
for i, im in enumerate(imlist):
    image = mpimg.imread(result + "/" + str(im, 'utf-8'))
    plt.title("search output %d" % (i + 1))
    plt.imshow(np.uint8(image))
    f = plt.gcf()  # 获取当前图像
    f.savefig(r'D:\pythonProject5\result\{}.jpg'.format(i),dpi=100)
    #f.clear()  # 释放内存
    plt.show()