[MUSIC] Now, these Watson custom classifiers are very good for what I thought are these two reasons. But the first one is when the visual inspection by humans itself is very time-consuming, difficult, unscalable, or very expensive, right? So that's, the image is gone. Come on, detect my computer. Give that a second. If we have remote maybe we could remove this date time setting. It's gone okay, very good. So the first point is when the visual inspection by humans itself is very time consuming. And the second part is when the process of training humans to become very, very good at detecting certain objects or certain things, or recognizing or classifying certain images is as well time consuming, difficult, unscalable or expensive. So you can think of when you go to the airport and someone has to manually inspect all of your x-rays of your baggage, or perhaps inspecting the tumors on an x-ray. It takes a lot of time to train these people to become experts at detecting strange things in their luggage, or strange things in your chest. So maybe these custom classifiers are a great way to use that. Here's another very different example. So the problem comes with videos and taking videos. We're often there with Instagram and taking a lot of our photos and everything like that. People already tag them quite well. In fact, I think everyone is overtaking hashtags and everything like that nowadays, Twitter and Instagram and everything like that. But what about videos, how do you go about taking videos? So what you can do with Watson is, and they've created this kind of prototype called OpenWhisk Dark Vision. What you do is you simply take a video and it's going to go through the video, and then it's going to take each of the frames within that video. And so what it can do now with this is you can, let's say you want to search for a cake in the video, and you don't know where it is in the video. So with these tags you can click on cake, and it will jump to that part in the video. So you can imagine how important this might be for security. So if you have a security tape, you want to jump to a point where something is happening, that can be tagged. Then maybe you can click on that tag itself and just jump directly to that point without having to watch the entire video from start to end. So this is kind of really one interesting example of how you can use Custom Classifiers in a very different way. Okay, so I want to go into the demo now. Again, we'll start off with Watson, and then afterwards we'll jump to the OpenCV. So I'll begin by showing you this. And so for this part there's no kind of hands-on part. There's no code, it's just all part of the the visual kind of GUI. And let me just open that up here. So to begin, if you want to take a look at these Watson API's beyond visual recognition, they're all here. You can type in Watson developer cloud or Watson APIs in Google, or you can type in cocl.us/watsonservices. And that will get you to the site, and you can take a look at all of these different services. There's also demos and everything in API documentation for everything as well. And if you'd like to learn a little bit more about each of these you can check these out. If you really want to be able to apply these to your own kind of industries, to your own work, please do check out the cognitive build course developed by, there he is, Jeff. Our Frenchman that, like Liz likes to say. So he is good at that to teach everyone how to build solutions using Watson. Okay, so by clicking on the visual recognition one you're lying on this page, assuming that you've logged into Bluemix. So you do need a Bluemix account in order to try using this visual recognition tool for free. And of course there is a free plan and there's a limit of the number of things that you can do per day, and after that it requires some price. It's very little, I don't know why this is in euros right now. [LAUGH] But it's like less than a cent for this or for that, all right? So you can go ahead and click on Create, and that will generate your service, this is your visual recognition service on Bluemix. And then you should see something like this. So this is my visual recognition service on my account, and I can click on, well, there's a couple of things I can click on. Service credentials is where I can get my API key, so that's really important for later on, and then the GUI, so there's a kind of interface it's called the tool, the visual recognition tool. So by opening that up, I should see my default classifiers. So here I've prepared some images, and you guys are all probably going to be skeptical about all these images, let me try to move this over here so I can kind of just drag it in. All right, so images to test. So a couple things here. I want to really kind of challenge Watson. And I tried these without kind of knowing what the answer would be. So let's try first default. Default kind of looks at any kind of object and tries to classify what kind of object that is. So here I'm uploading this image of sushi. Maybe I can increase the size of this, no? It’s okay, let that run a second. And in this second classifier, I can also upload something else. So let’s try uploading this taco. There’s also one more image here that I’d also like to try. Let me just try to zoom in here. So this is actually an image of a taco, and it’s kind of hard to tell, right? It’s kind to hard to tell that it’s a taco, but Watson can actually classify that quite well within this food recognition tool here. So let that run for a second. And for faces, well, we can add in. I tried Mount Rushmore, it doesn't work. Mount Rushmore, let's try to see if that works. [MUSIC]
