Tensorflow Serving with Tensorflow Slim Models
Overview
Here comes the interesting part!
In previous section, we trained, built and served models such as mnist, inception-v3 that comes with the tensorflow serving. In this section, however, we will build and serve customized models.
To keep it simple and clear, we will use pre-defined ./tf_models/research/slim/nets/inception-v4 and ./tf_models/research/slim/nets/inception-resnet-v2 models as examples.
This approach should works with all models defined in ./tf_models/research/slim/nets, such as inception, inception-resnet, resnet, vgg, and mobilenet. A full list of models and versions can be found at https://github.com/tensorflow/models/tree/master/research/slim.
Let's start!