[MUSIC] The process that I'm going to talk about here it doesn't involve any special hardware it's called photogrammetry. And you might guess that means it uses photos so, measuring distances within photographs is photogrammetry. And modern software has allowed us to measure millions of distances between objects inside photos which lets us create entire 3D reconstructions, 3D meshes, from just photographs. So what that means is that you can make really high resolution 3D models with no special hardware whatsoever. The way that the computer creates that 3D reconstruction from photographs, is it compares one photograph to the next one, and looks for common features. So, we're going to be photographing this berry tart. And from one photograph to the next, the computer's going to be figuring out. Well, here's a shape in this photograph, that's the same as this shape in the other photograph. So it's called feature detection and it's just a lot of edge finding algorithms going on, to look at one photograph and find the same shape, but a little bit distorted, you're coming at it from another angle. And defining that same shape in the next photograph, and doing that for dozens of photographs. So the process, no matter what software you use is going to be taking your camera and taking pictures of every angle, and you can kind of imagine that you're just capturing two-dimensional data but from every angle. From that set of photographs the computer will be able to determine what that object is shaped like in real 3D space. So you can use whatever camera you have on hand whether it's a compact camera with a zoom, or if it's your cellphone camera. Both work great. The most important part of taking that set of photographs is to get everything in focus, and to not change anything about the lighting, anything about the background, or even anything about your zoom. The object has to remain as static as possible and something you might not think about intuitively is that when you zoom in and out with the camera. The way that those lenses are moving forward and back from each other to zoom in, actually changes the shape of the object as far as the camera is concerned. Cellphone cameras have a neat trick that allow them to be actually really useful for this process. Is that it's important for everything to be in focus, and cellphone cameras have such a small sensor and such a small set of lenses that they're sort of built to always get everything in focus. You can use these photographs in a lot of different types of software. So if you can just look up photogrammetry software, I'll show you, the one I'm familiar with is Agisoft. Autodesk also provides a 123D Catch. So I prefer to just take all the pictures first and then decide what software I want to use later. So my strategy is to kind of of pick one side to start off with. Start off with a kind of top down view. You want to fill a lot of the frame with the object. If I was to fill the entire frame, then I would actually run into some of it being blurry and some of it being sharp, so I do want to get to a range where the entire object is in focus, so this looks pretty good. So I'll start from top down and the shooting strategy actually works really well if you think of the way that humans do depth sensing, we've got our two eyeballs, a certain distance apart, and that helps us do depth perception. So when I shoot this raspberry tart, I'm going to take one picture, shift the camera like just an inch to the right and take another picture. And so those are called binocular photographs. Sort of taking one picture right after the other to have a binocular set. So it will really help the computer determine what photographs go together, and just give a little bit of depth to each set of photographs. Putting this into 60 degree increments, so you just take one, two, three, four, five, six. If you just imagine a hexagon, that's a pretty good results. If you want to get a really high resolution, then you can do like 10 degree increments, and do, that'd be 36 angles. And the more you do that, the higher resolution you can get, but I will also greatly increase the processing time. So once we take all these photographs, we have to look at the computer, chew in all that data. The more pictures you take, the longer it will take the process. The camera at one point or another sees every surface that's a part of this. So this berries are really funny shapes that have a lot of different sides, I want to even get a little bit underneath them so that the camera can see the underneath the bottom part of these berries. Picture there, shift the camera a little bit, picture there. It might be tempting to rotate the object but while that worked with some of the automatic 3D scanning. Especially, the connect can deal with objects spinning around. With photogrammetry it uses the object itself, but will it'll also will use background information to match photographs. So you want to keep the entire scene as static as possible. So I can go to my photos and just show kind of what that entire photo set looked like. Dozens and dozens of pictures of this object from every angle. We can take these photographs and pump them through our choice of software. I'll be demonstrating Agisoft PhotoScan, which is a free to try software. You can download the demo and for 30 days, it's unrestricted. You can try it out with whatever you want. Another interesting option, probably the easiest option, you just download the app and start shooting is called 123D Catch. So, it's an AutoDesk product and it runs on Android and iOS and all that, so I'll demonstrate how automatic it is because it is a little bit more trouble free. The software walks you through it so I definitely recommend trying it out and if we take a look at it we can kind of see that here, this shows me where I am in space. So it's using my accelerometer to kind of track my phone. And so if I set it from the side it kind of switches to this set of photographs. So all these little boxes are demonstrating photographs that you should take. And the blue are photographs that I've already taken. So I'll shift it here it says I need to take a photograph from this angle. Click. I can shift down. It says I should take a photograph from this angle. But it's important to not like only listen to this because it will say, yeah, that's a good place to take a picture from but, of course, my object is out of frame. So of course, keep your object in the frame. Just use these as a guide, kind of a reminder of what pictures you've already taken. Doing one ring around from the side, and a smaller ring around from the top. Once you take a picture, it's not set and stone, you can always move back and save over it. So I can move to this angle, and kind of take this picture and say, you know what? The one I took just now was a little bit out of focus, so let me just take that picture again. And I'll just finish this up as I go all the way around. There. And then, so now I can see that I have my entire ring kind of coated blue, so I can hit my done checkmark. So it says, review your photos. And I can maybe see that a few of these, okay, the whole object's there. But it yeah, so you can kind of go over and check. You know what? I need to reshoot this one, because my object is not all the way in the frame. So I'll hit reshoot, and it'll tell me kind of what side it's coming from. But now I can kind of get the whole object in the frame. And that's a better picture, okay. All right, so I can submit now. But this is another software strategy where you're offloading the job to someone else's computer. So that's cloud computing, right. And I'll sign into my account so it requires a fast Internet connection, it's going to upload maybe 100 megabytes or more of data. With any of these technologies, the result of the 3D scanning is not going to be a file that's ready for the 3D printer. There's one last step to do before you can hit prints. With even a high quality reconstruction of my toy train it's just a surface of one side of it, so this is just a shell. If I send this to the 3D printer, it wouldn't know what to do with it, same thing with my adding machine. It's a really beautiful 3D model, but it has no bottom. So, again, there's no volume here, it's just a surface. And if I was to send this to the 3D printer, it wouldn't know where to start. So there's one last step before sending it to a printer, and that's modifying the file to create a water type volume which is to say a solid object that the computer could make sense of when you ask it to print with plastic, whatever material you're going to working with. So the adding machine, it would be easy. I just need to open it in a program and seal the bottom up. The steam model might take a little bit more creativity. I might duplicate it and reflect it but there's a step of modifying these scans before you can send it to a printer. So we'll show you a couple of different strategies for that in the next video. [MUSIC] [SOUND]
