Training TensorFlow Object Detection Models

This picture shows the track with two cars and a phone and the iOS app which draws rectangles around the objects.

1) Development Environment Setup

First you need to download the trained MobileNet model which is an optimized model for mobile devices. Rather than training a new model from scratch, transfer learning is used. Basically the last layer of the neural network is replaced with your own objects.

2) Labelling of Images

From what I’ve read these are some best practices how to create the training data:

3) Training of the Model

After the training a frozen graph of the model needs to be created. The repo contains a script and a Docker container to do this.

4) Usage of the Model in Notebooks and Apps

Connecting the Cars and the iOS App to the Watson IoT Platform

Here is a diagram of the high level architecture:

Related posts:

A safeword is a word or signal that ends the BDSM scene immediately. It can be any word that you don’t expect to be using during the scene. Common stopwords are “red” or a word you don’t use by…

This is my first test story.. “Test Story 1” is published by Ian Stewart.

This is a short extract from my forthcoming book “Save Democracy — Abolish Voting” which will be published by The Democratic Society on the 16th November 2017 [launch event]. A ballot paper is…