Difficulty: beginner
Estimated Time: 10 minutes

TensorFlow is the platform enabling building deep Neural Network architectures and performing Deep Learning. One of the achievements was tackling the challenge for ImageNet, the well known image database.

This tutorial introduces how to use pretrained model based on Inception V3 architecture to recognise new images.


You've completed TensorFlow Inception scenario.

You've learned how to:

  • Serve the model using docker image
  • Connect to the server
  • Query the server to classify new examples

TensorFlow Inception

Step 1 of 5

Introduction to Inception

TensorFlow proved to be very successful with many AI tasks. One of the most popular in computer vision in the ImageNet exercise. The exercise is to test the model against the database if the images and their proper labels.

Inception in Tensorflow is a project that showcases the training of the Inception V3 architecture.

Inception V3 Architecture

Once trained it can be used to classify new examples. To make things easier instead of building the model, packaging and maintaining it yourself, one can build the docker image. Katacoda offers the katacoda/tensorflow_serving so there is no need for us to create one for the examples.