FB Pixel

Google Cloud Vision | Behind The Scenes of Artificial Intelligence

Google Cloud Vision drives much of the AI photo software we see available today. In this video, I walk through what Google Cloud Vision is doing behind the scenes and how it can impact photography software you can buy and use.

Transcription was done by Descript‘s automated transcription services which means it’s an AI-generated transcript. The transcript may contain spelling, grammar, and other errors, and is not a substitute for watching the video.

Artificial intelligence, machine learning. These things are getting smarter and smarter, and it’s thanks to companies like Google and IBM and all these companies that these large companies that are creating these platforms to really experiment with Ai. We’re seeing so many different types of AI being used for photographers on a regular basis. Thanks to companies like sky loom and companies like Topaz who are pushing the limits of what’s possible right now in photography with Ai. But today I want to take a step back and show you some of the behind the scenes of what AI is doing, and I want to show you from Google cloud’s vision platform, what it can do in recognizing what’s inside of an image. Now, if you watched my video on colorize that Sg, you will recognize these two photos that I have used in that video to colorize old black and white photos and I’m going to use that in this video as well.

Those two photos to show you what Google clouds can do behind the scenes, but before we get started, please click the subscribe button below. Right now. I publish new videos every Monday and Thursday whenever possible. You don’t want to miss it and karate kick the notification bell so that you get notified when the next video goes live. And a quick thank you to Matt old I warranty. I partner with them for these videos and I get a Mac ward, a warranty on every product I own, including the computer and the camera that I’m recording on right now. Check out map auto warranty if you need coverage for your equipment, including accidental damage. So this is the Google Cloud Vision API action. This is the sort of the behind the scene. This is what happens when it’s companies like skyline utilize Google cloud. Look, I don’t, I don’t know if skyline would actually using Google cloud vision or if they’re using their own AI.

I don’t know. Uh, but this is what it looks like behind the scenes from a coding standpoint. What the software will be looking at. I’m going to just drag in an image. This is an image of my grandparent’s house from the forties or fifties I don’t know when it actually is. And you can see that it’s actually recognizing some of the objects. It’s recognizing a car. It’s recognizing another car and it’s recognizing a house. It’s 94% positive that this one has a car and it’s 69% positive that this is a car and a 61% positive that this is a house. I don’t know why it’s not 100% positive, but hey, you know, it did okay. It still recognizes that that’s what it is. We can move on to labels and you can see that it’s actually labeling some objects as well. These are basically label, these are like keyword that could be used.

Um, and you can see again, all these are over 54% positive. I’m led, it’s a sedan. It’s a vintage car, it’s a suburb. Um, classic car, residential area, neighborhood. It’s a snapshot. It’s giving all these keywords that could be used. Um, and think about this. This is all stuff that Google is now using in their search as well, which means that these keywords could in theory bring up this photo as well in a Google image search depending on what you searched for. Right? All right. Now if you go to web, again, here are some searches that you can do based on those keywords. We go to properties and you can see it’s giving me the color tones, the actual hex values, the RGB values and how often it’s being used. And of course I hit, this was a color photo. It would be showing it to you with the colors instead of just the grayscale colors.

So you can actually get you actually she pick out the color tone being used in a photo just from dragging an image right in here, which is a pretty smart thing. It also gives you some other aspect ratios and things that you know, uh, may not be that useful in this case. And then safe search, it tells you how safe this would be for, you know, is it adult photo, is that medical stuff like that. Is there violence? So it tells you that as well, which also can tell you how a photo might show up in search results. So if you are in a good word photographer, uh, and somebody has a McCount that has safe search turned on, maybe you may not come up. And the way to find out, it’s a look over here at the safe search values for the, for that photo.

And if you really want to, you can actually download a Jason file with all of this information in it. That’s not probably not very useful for most people, but you never know for developers it might be useful as a demo data. All right, so I’m going to reset this and we’re going to do one more photo. I’m going to do the photo of from Dachau of my wife’s grandparents grandfather being liberated from the concentration camp. And you’re going to see here that now it has a new, a new thing called faces because this is, there’s people in it. It’s actually going to show me that in a, show me how many people are happy, how many people are wearing headwear and, and all these different things. And you can see that there’s a lot of people who are happy because they were liberated. It was a bunch of people wearing headwear because they probably, they probably made it themselves.

Um, there’s, there’s a lot, there’s some people that are angry is some sorrow. There’s some surprise. Um, it’s there. It’s not a lot, but it’s there. Uh, there’s a lot more joy than there is anything else and that makes sense, doesn’t it? They were just liberated from a concentration camp. Um, you can see the next phase. It’s good basic going one by one and it’s showing you, you know, again, headwear um, joy headwear and it’s going face by face of all faces it can detect. Right? And that’s really, really cool thing. It’s telling you what the facial expressions are. We go to object and you’re going to see men, men over men of romance. So many men are being detected. No women have been detected at this one. And of course a flag has been detected. This is a homemade flag with maid, uh, to welcome to the American troops.

We go to label. And again, you’re seeing all these things that are, that are, uh, that are recognized troops, rebellion, military officer, infantry. Now these are not military. These are, you know, people who were kept as basically a slaves in the concentration camp. Um, but uh, you know, it’s, it’s not perfectly accurate but it’s got a good idea of what it actually is. We go to web and these are all different things that might be relevant search terms. And you can actually click in and click on these and bring up the actual search and you can see, uh, where this introduction shows up. So this, this image is in the Holocaust museum. So it shows up a lot of places, shows, evening shows, it even shows up on Getty images and, and so on. There’s a lot of places where this can be found. And then again, it’s going to show you a dominant colors.

If I was too, and I will actually drop in the color version of this so you can see the colors that it will pick a, but as you saw earlier with the other photo, it’s showing me the grayscale and then save search. It’s showing me, uh, that’s a little bit of violence right, is it does recognize it as they holocausts photo. So it is going to show a little bit of violence here, but otherwise safe search is okay. Now let’s reset this. I’m going to drop in the color just so you can see that it’s also picking up the colors, uh, in the um, with the hex values. I’m going to go to a, I think it was one was we have a properties. So here it’s picking up the colors. You can see the colors that is choosing, giving me the hex values and the RGB values and how much it’s much that is actually you.

So like this part of this guy is 53%. This part of the sky is 21. This part of the sky is 9% skin tones. You’re talking, you know, 2% and so on until all this is right there, which is really cool. So that is Google cloud vision. That is the behind the scenes of what Google is doing for anybody, for any developers, any software using Google cloud vision in their tools. This is sort of what data is given to that software. And I think it’s so cool to see how, how accurate and how much value can be provided from this artificial intelligence to software. Because that means that for example, Skylar rooms, a accent AI can actually pull in the sky because he literally recognizes the sky. It legally recognizes the sky and the sky is tones and the colors and things like that. It can literally pull in details from clothing or from the dense in cars because it recognizes those objects as well. So I hope you enjoy this. Uh, this, this other video on artificial intelligence in photography. I’ve got an entire series of, of, of videos that are about artificial intelligence in photography and photography software and whatnot. So it’d be sure to check that out. I really hope that you enjoy this video. And I’ll see you in the next one.

By Scott

If you liked this post, you will definitely enjoy our others. Subscribe to the blog via Email.

Leave a comment

Your email address will not be published. Required fields are marked *

Share
Share
Pin
Email
WhatsApp
More