Training your own object detector with TensorFlow's Object Detection API
It's no secret that modern object detectors rank among the most complex and challenging architectures to implement and get it right! However, that doesn't mean we can't take advantage of the most recent advancements in this domain in order to train object detectors on our own datasets. How?, you ask. Enter TensorFlow's Object Detection API!
In this recipe, we'll install this API, prepare a custom dataset for training, tweak a couple of configuration files, and use the resulting model to localize objects on test images. This recipe is a bit different from the ones you've worked on so far, because we'll be switching back and forth between Python and the command line.
Are you ready? Then let's get started.
Getting ready
There are several dependencies we need to install for this recipe to work. Let's begin with the most important one: the TensorFlow Object...
