image-recognition

invoke a python-trained tensorflow model from java programs,Real-time image recognition,and deploy the applications to springcloud dataflow local server

this is my Graduation Project.

1. use spring cloud task,run the retrain.bat

you can train your own data by run the retrain.bat runs the tensorflow-master/tensorflow/examples/image_retraining/retrain.py and uses the inception-v3 model #inception-2015-12-05.tgz

After execution,under the classpath,two files are generated,output_graph.pb and output_labels.txt.

Of course you can modfiy the application.properties to change the path

2. start source,processor,sink applications

you can modfiy their application.properties according to your programs

spring cloud stream source application: put picture in the configuration folder spring cloud stream processor application: invoke the python-trained tensorflow model,and get the image recognition result spring cloud stream sink application: just a simple ui used freemarker to show the result

3. now we can package the sateams,source,processor,sink

$ mvn clean package

4. start the data flow server

5. we can use the Data Flow Shell to deploy the sateams and task.

The Data Flow Shell is a client for the Data Flow Server. The shell allows us to perform the DSL command needed to interact with the server

$ java - jar spring-cloud-dataflow-shell-1.2.3.RELEASE.jar

dataflow>app register --name source --type source --uri maven://com.tse.graduation:springcloud-stream-rabbit-source:jar:0.0.1-SNAPSHOT

dataflow>app register --name processor --type processor --uri maven://com.tse.graduation:springcloud-stream-rabbit-processor:jar:0.0.1-SNAPSHOT

dataflow>app register --name sink --type sink --uri maven://com.tse.graduation:springcloud-stream-rabbit-sink:jar:0.0.1-SNAPSHOT

dataflow>stream create --definition "source --server.port=8093|processor --server.port=8094|sink --server.port=8095" --name mySimpleImageRecognition

dataflow>stream deploy --name ss --properties "app.source.spring.cloud.stream.bindings.output.destination=springcloudstream, app.processor.spring.cloud.stream.bindings.input.destination=springcloudstream, app.processor.spring.cloud.stream.bindings.output.destination=springcloudstreamsink, app.sink.spring.cloud.stream.bindings.input.destination=springcloudstreamsink, app..spring.cloud.stream.bindings.output.binder=rabbit1, app..spring.cloud.stream.bindings.input.binder=rabbit1"

Note:deploy the sateams,app.*.spring.cloud.stream.bindings.output.destination should be consistent with your programs

6. the you can see the logs and visit the sink application's ip:port/index to check the result