Tensorflow Serving with Tensorflow Slim Models
Overview
In previous section, we built and served several slim models, including inception-v4 and inception-resnet-v2. These native Tensorflow Serving services uses gRPC and protobuf as communcation protocal. A similiar service uses REST and JSON could also be beneifical for general users.
In this section, we create a REST client slim_unified_client_rest.py, corresponding to the gRPC client
slim_unified_client.py, and demonstrate how to serve this API using uWSGI and Nginx.
In case no interest in uWSGI and Nginx, skip to section about docker image directly. The docker image section shows how to use the gRPC and REST service from docker image, rather than how to build these service.
Let's start!