Using object detection models in iOS

In the previous chapter, we showed you how to use the TensorFlow-experimental pod to quickly add TensorFlow to your iOS app. The TensorFlow-experimental pod works fine for models such as Inception and MobileNet or their retrained models. But if you use the TensorFlow-experimental pod, at least as of this writing (January 2018), with the SSD_MobileNet model, you’re likely to get the following error message when loading the ssd_mobilenet graph file:

Could not create TensorFlow Graph: Not found: Op type not registered 'NonMaxSuppressionV2'

Unless the TensorFlow-experimental pod gets updated to include op not registered here, the only way to fix these problems is to create the custom TensorFlow iOS library by building it from the TensorFlow source, and that's why we showed you in Chapter 1, Getting Started with Mobile TensorFlow, how to get and set up TensorFlow from its source. Let’s look at the steps to build your own TensorFlow iOS library and use it to create a new iOS app with TensorFlow support.