构建应用深层特征的图像检索系统

从GraphLab Create 库里面导入一个深度学习的模型

deep_learning_model=graphlab.load_model(‘http://s3.amazonaws.com/GraphLab-Datasets/deeplearning/imagenet_model_iter45‘)

接着,利用deep_learning_model从我们定义的image_train中去抽取深度特征

deep_learning_model.extract_features(image_train)

执行完毕之后,

我们的深层特征也是给求出来了

knn_model=graphlab.nearest_neighbors.create(image_train,features=[‘deep_features’],label=’id’)

通过运用深度特征的图像检索模型找到相似图像

通过运用深度特征的图像检索模型找到相似图像¶
In [8]:

cat =image_train[18:19]
In [14]:

graphlab.canvas.set_target('browser')
In [15]:

cat['image'].show()
Canvas is accessible via web browser at the URL: http://localhost:55693/index.html
Opening Canvas in default web browser.
In [11]:

cat['image'].show()
开始检索
In [17]:

knn_model.query(cat)
knn_model.query(cat)
Starting pairwise querying.
+--------------+---------+-------------+--------------+
| Query points | # Pairs | % Complete. | Elapsed Time |
+--------------+---------+-------------+--------------+
| 0            | 1       | 0.0498753   | 127.073ms    |
| Done         |         | 100         | 311.842ms    |
+--------------+---------+-------------+--------------+
Out[17]:
query_label reference_label distance    rank
0   384 0.0 1
0   6910    36.9403137951   2
0   39777   38.4634888975   3
0   36870   39.7559623119   4
0   41734   39.7866014148   5
[5 rows x 4 columns]
通过 reference_label关联id得到images
In [18]:

def get_images_from_id (query_result):
    return image_train.filter_by(query_result['reference_label'],'id')
通过上面的函数得到猫的邻居
In [19]:

cat_neighbors
cat_neighbors=get_images_from_id(knn_model.query(cat))
Starting pairwise querying.
+--------------+---------+-------------+--------------+
| Query points | # Pairs | % Complete. | Elapsed Time |
+--------------+---------+-------------+--------------+
| 0            | 1       | 0.0498753   | 18.013ms     |
| Done         |         | 100         | 241.681ms    |
+--------------+---------+-------------+--------------+
In [20]:

cat_neighbors['image'].show()
​
Canvas is updated and available in a tab in the default browser.

接下来找到和红色小轿车相似的图像

代码:

car=image_train[8:9]
car['image'].show()
get_images_from_id(knn_model.query(car))['image'].show()

构建一个lambda函数来方便寻找

返回image的函数

def get_images_from_id (query_result):
    return image_train.filter_by(query_result['reference_label'],'id')

show_neighbors=lambda x: get_images_from_id(knn_model.query(image_train[x:x+1]))[‘image’].show()