An easy way to add machine learning to your applications, is to take advantage of pre-trained models. These are off-the-shelf solutions. Cases that you don't need to build your own models. Aucnet is the largest real-time car auction service in Japan. Serving over 30,000 dealers and running auctions worth nearly 4 billion dollars a year. And the way it used to work was that car dealers would take multiple photos of each used car to sell, upload them to the service, and need to specify what model of the car and what part of the car for every photo. It's a time consuming task for the dealers to do across thousands of photos every day. And now, the new machine-learning system can detect a model number of the car at high accuracy. It can also show the estimated price range for each model and recognizes what part of the car is being photographed. With this system, the car dealers just drag and drop a bunch of unclassified photos, and then check if the model and parts are classified with the system correctly. So let's see how the Aucnet website works. So here's an example of what the website looks like, you can basically go ahead and upload images of a car. So, at this point we're uploading some photographs that we've taken of a car. And what we're going to see is that the system is going to take all of these images and catalog them. It's going to basically find which photographs are from the front of a car, which are of the side of the car etc. It's also going to identify the brand of the car and use those as inputs to come up with an evaluation of what this car might be worth. So here it is, you see that we uploaded a photograph of a truck, and it said this is the front of the truck, here is a right side of the truck, the right side front. I didn't upload the photograph of the rear or of the left side rear, but I did have one image of the front tire and it's all there. So I basically went ahead and put in all of those images in the appropriate slots and set at first guess 96 percent confident that this is a Toyota Land Cruiser. Just notice how much easier this is. Rather than the human user having to upload a whole bunch of information, typing a whole bunch of information into a form. All that they're doing now is uploading a bunch of photographs, and the system is cataloging them, identifying what the model of the car is, and being able to go ahead and get them further along in the process of listing their car for sale.