Segmenting images using Mask-RCNN and TensorFlow Hub

Mask-RCNN is a state-of-the-art architecture for object detection. However, as its name suggests, it's also excellent at performing image segmentation. In this recipe, we'll leverage an implementation of Mask-RCNN hosted in TensorFlow Hub (TFHub) that has been trained on the gargantuan COCO dataset. This will help us perform out-of-the-box object detection and image segmentation.

Getting ready

First, we must install Pillow and TFHub, as follows:

$> pip install Pillow tensorflow-hub

We also need to install the TensorFlow Object Detection API since it contains a series of convenient visualization tools that'll come in handy for looking at the bounding boxes and segmentation masks. First, cd to a location of your preference and clone the tensorflow/models repository:

$> git clone –-depth 1 https://github.com/tensorflow/models

Next, install the TensorFlow Object Detection API, like this:

...