keras search engine

Keras-Search-Engine

KERAS中實現的簡單文檔和圖像搜索引擎

特徵

• 使用手套字嵌入文檔級別編碼
• 使用手套嵌入支持全文搜索
• 支持文檔通過Web API索引
• 支持圖像搜索
• 支持圖像通過Web API索引
• 使用LSTM編碼器與手套嵌入的編碼器進行自動編碼

用法

運行Web API服務器

goto keras_search_engine_web目錄並運行以下命令：

python flaskr.py

現在將瀏覽器導航到http：// localhost：5000，您可以嘗試一下

• 故事書搜索（用戶通過keras_search_engine_train/data/texts在故事書上輸入搜索查詢）
• 圖像搜索（用戶將圖像上傳到keras_search_engine_train/data/images中存儲的圖像之間搜索圖像）

調用Web API索引和搜索文本文檔

在運行Web API服務器後，您可以通過調用以下Web API來索引新文檔：

http://localhost:5000/index_text

例如，以下是curl命令調用Web API索引一些文檔：

curl -H ' Content-Type application/json ' -X POST -d ' {"doc":"Whether you think that you can, or that you can ' t, you are usually right. " }' http://localhost:5000/index_text
curl -H 'Content-Type application/json' -X POST -d '{ " doc " : " Try to learn something about everything and everything about something. " }' http://localhost:5000/index_text
curl -H 'Content-Type application/json' -X POST -d '{ " doc " : " You can avoid reality, but you cannot avoid the consequences of avoiding reality. " }' http://localhost:5000/index_text
curl -H 'Content-Type application/json' -X POST -d '{ " doc " : " A mathematician is a device for turning coffee into theorems. " }' http://localhost:5000/index_text
curl -H 'Content-Type application/json' -X POST -d '{ " doc " : " In theory, there is no difference between theory and practice. But in practice, there is. " }' http://localhost:5000/index_text
curl -H 'Content-Type application/json' -X POST -d '{ " doc " : " I find that the harder I work, the more luck I seem to have. " }' http://localhost:5000/index_text

要使用Web API查詢，您可以調用以下Web API：

curl -H ' Content-Type application/json ' -X POST -d ' {"query":"mathematician and coffee", "limit": 3, "model": "glove"} ' http://localhost:5000/search_text

調用Web API索引和搜索圖像

在運行Web API服務器後，您可以通過郵政請求調用以下Web API來索引新圖像：

http://localhost:5000/index_image

您可以通過調用以下Web API發布請求來查詢類似的圖像：

http://localhost:5000/search_image/10

其中10是返回的圖像數量的限制

使用SearchEngineClient

keras_search_engine_client中還有一個searchengineclient類，示例代碼看起來像：

 from keras_search_engine_client . search_engine_client import SearchEngineClient
client = SearchEngineClient ()

# text indexing and search
doc_count = client . doc_count ()
if doc_count < 4 :
    client . index_text ( 'Whether you think that you can, or that you can.' )
    client . index_text ( 'Try to learn something about everything and everything about something.' )
    client . index_text ( 'You can avoid reality, but you cannot avoid the consequences of avoiding reality.' )
    client . index_text ( 'A mathematician is a device for turning coffee into theorems.' )
    client . index_text ( 'In theory, there is no difference between theory and practice. But in practice, there is.' )
    client . index_text ( 'I find that the harder I work, the more luck I seem to have.' )
client . search_text ( query = 'mathematician and coffee' , limit = 3 , model = 'glove' )
client . search_text ( query = 'mathematician and coffee' , limit = 3 , model = 'doc-encoder' )

# image indexing and search
client . index_image ( './images/Pokemon7.png' )
client . search_image ( './images/Pokemon1.jpg' , 6 )

配置以在Windows上的GPU上運行

• 步驟1：將TensorFlow更改為unignts.txt中的TensorFlow-GPU並安裝TensorFlow-GPU
• 步驟2：下載並安裝CUDA®工具包9.0（請注意，當前CUDA®工具包9.1尚未由TensorFlow支持，因此您應該下載CUDA®工具包9.0）
• 步驟3：為Cuda@ Toolkit 9.0下載並解開Cudnn 7.4，然後將UNZPRECT DICTIRORY的bin文件夾添加到Windows環境的$路徑中