3D Photo Scanning for Beginners | Eric Hayes | Skillshare

3D Photo Scanning for Beginners

Eric Hayes, Filmmaker & Engineer

3D Photo Scanning for Beginners

Eric Hayes, Filmmaker & Engineer

Play Speed
  • 0.5x
  • 1x (Normal)
  • 1.25x
  • 1.5x
  • 2x
7 Lessons (29m)
    • 1. Introduction

    • 2. How Does It Work?

    • 3. Pros and Cons

    • 4. Software

    • 5. 3D Scan an Object

    • 6. 3D Scan an Environment

    • 7. Conclusion

  • --
  • Beginner level
  • Intermediate level
  • Advanced level
  • All levels
  • Beg/Int level
  • Int/Adv level

Community Generated

The level is determined by a majority opinion of students who have reviewed this class. The teacher's recommendation is shown until at least 5 student responses are collected.





About This Class

Turn photos taken with your phone into a 3D model!

In this class we'll cover the basics of 3D photo scanning (also known as photogrammetry). Using only pictures taken with your phone and accessible software, we will turn real world objects into 3D models.

This is a powerful and versatile technique that's useful for animation, visual effects, game development, 3D printing, architecture, and more! It's used by the film and game industries to create assets, by museums to create archival scans of artifacts, and even in medicine for medical imaging. Whether you're a CG artist, looking for an easy way to produce 3D scans, or just looking to try something fun, this class is for you!

By the end of this class you'll have the knowledge you need to create 3D scans of your own.


Some helpful resources and guides:

Meet Your Teacher

Teacher Profile Image

Eric Hayes

Filmmaker & Engineer


Hi there, I'm a filmmaker and engineer with a passionate for storytelling. I've been fortunate enough to work at studios like DreamWorks Animation, Sony Pictures, and Weta Digital. I'm excited to share some of skills and tools you can use to be creative and make cool stuff!

See full profile

Class Ratings

Expectations Met?
  • Exceeded!
  • Yes
  • Somewhat
  • Not really
Reviews Archive

In October 2018, we updated our review system to improve the way we collect feedback. Below are the reviews written before that update.

Your creative journey starts here.

  • Unlimited access to every class
  • Supportive online creative community
  • Learn offline with Skillshare’s app

Why Join Skillshare?

Take award-winning Skillshare Original Classes

Each class has short lessons, hands-on projects

Your membership supports Skillshare teachers

Learn From Anywhere

Take classes on the go with the Skillshare app. Stream or download to watch on the plane, the subway, or wherever you learn best.



1. Introduction: Hi, My name's Eric. I work in the film industry in animation and visual effects. And today I'm gonna teach you about something really useful, Which is how you can create a three d scan of something and turn it into a three D model using Onley pictures taken with your phone and some free software. Now, I'm sure you're wondering. Is this an Oscar? And in fact, it's the even more prestigious runner up prize that I got for a film I made in high school . So, uh, yeah, but since then, I've had the chance to work in a few different studios on some pretty cool projects. So I'm gonna share some secrets. Creating three D models for a scene or four game can be pretty difficult, because you need to know how to model it. You have to texture it. Um, and if you don't know how to do that, you're usually going online, and you might have to buy something that someone else has made and that costs money. I'm gonna show you a really easy way to do it yourself That can pretty quickly get you some pretty good results. This is really useful for anyone who is trying to do animation or game design or visual effects, or even if you just want to make cool gifts for people. If you've ever seen those many statues of people that are scanned and printed, you could do the same thing using this technique and then printing the model in a three D printer or something like that. In fact, once you've got these models that you've made that you're really proud of, you can even put them online for other people to buy or download. So I hope you find this really useful because it's amazing what it can do. 2. How Does It Work?: So how does this actually work? How can you take two D pictures and turn them into a three D model before we jump into talking about how to take your pictures and what kind of software you can use? I just want to briefly go over how this actually works. There's a bunch of different software that does this, but most of it is using the same kind of strategies to do it, and we're just gonna quickly talk about what those are. When we start talking about the pros and cons of how this works and what it's good for and what is not good for, I think it's gonna make a lot more sense. So I'm just quickly going to go through the four steps that most pieces of software are using to turn pictures into a three D model. Just for this example. Let's imagine this box is the object we started with, So it's just a simple box and its got a hole in the side for the handles. We've taken a bunch of pictures in slightly different angles, and now we're feeding it into the software, so the first step is going to be to try to figure out where those cameras were when it took those pictures. And the way it's gonna do it is just gonna look at each picture and it's going to start finding features in it. Different points like, for example, that little corner where the handle is. If you can imagine zooming in to the pixels, there's gonna be a dark spot on a light bit, and it's going to say, OK, there's a point and it's gonna do this a handful of times throughout the image, and it's gonna do the same thing for each of the images in your collection that is gonna try to do some pattern matching to say which one of these points matches the point in a different photo, and it's gonna do this over and over again until it has thes matches. And then when it knows where the same point is from a different angle, it's gonna work backwards and triangulate where the camera was, what it took this picture. So at the end of the first step, we took these images and we're gonna turn it into this really sparse cloud of points that corresponds to these matches and we're gonna know where the cameras were when it took these photos. In the second step, we're just gonna take this cloud of points and we're going to refine it and make it more detailed. So if you can imagine, we've got our loose cloud of points and we know where the cameras were when it took all these pictures is just gonna go back through now and make this a lot more detailed with a lot more points. Looks something like that. Say so in the third step, now that we have this cloud of points, we're gonna turn these points into an actual model. And really all that's doing is just connecting the dots and turning a point cloud into a mesh. And once it has this mesh, now we've got a three D representation of the object we started with. And in this first or second step, we might know the colors of each of these points. But we're going to try to create a more detailed texture. So in the fourth and final step, we're gonna take this mesh recreated and using the photos that we started with. We're gonna take the camera positions and project those photos back onto the mesh, and we're gonna blend those together to create a more detailed texture that gives you a much better representation of what this looks like. So in the first step, we took a bunch of photos in and after each of these steps, By the end of it, we have a textured three D model. 3. Pros and Cons: now that we kind of have an idea how this works, let's quickly go over what it's good for and what it's not so good for. So the good news is this is really versatile, and it's something that's used all over the place. So it's something you'll see a lot in video games for environments, whether it's little details on rocks or tree trunks or bigger areas. Even, um, it's really useful for those sort of organic types of shapes and areas, because when you have those different points of view in those pictures, it's pretty easy to find those points we're talking about and match them together. Um, and the textures really help with that, too. It's also really useful for things like props. So if you have something you don't want to recreate on the computer, you can probably just take it and take a bunch of different pictures from different angles and reconstruct it that way. So where should we not use this? There's a few situations where this reconstruction technique does struggle a little bit, and one of them is for anything that transparent. So, like this glass, for example, it's gonna have a hard time, and the reason for that is if we remember back to the steps we looked at, it's important for the computer to be able to look at your photos and find those points that it can match between images. And the problem with a transparent surface like this is if you take a picture from one angle and then you take another picture from a different angle. Those points on the surface because it's transparent don't look the same from those different angles. So it's really difficult to try to match those up, and because of that, you're gonna have a hard time reconstructing it. The same thing goes for really shiny surfaces for the same reason, because for something that shiny when you look at one angle, it's not gonna look the same as if you look at it from a different angle. Another place this can struggle a little bit is with really fine detail, So imagine, like for or grass something where there's a lot of intricate pieces. Um, it's only gonna work if you have a lot of really good images, and that's gonna be really difficult to do so generally, this works better for a larger kind of block here, more rough shapes, and the more detail that gets, the harder it's gonna be you to get a good reconstruction. 4. Software: There's actually quite a lot of software at the table to do this. I'm going to include a list of some of the software I know about. But there's a lot out there. If you're working on Windows, there's actually quite a lot more pieces of software I found eso If you're on a Mac like me , there's fewer options, but there's still a couple good ones. Just to be clear, I'm not being sponsored or paid to promote any of these brands. But I'm just gonna tell you about software that I've used and that I, like. All of these programs are actually pretty similar, and you'll see when we start to go through that, they're not that complicated. The main piece of software I'm gonna use today is called Photo Stand, or I think it might have been renamed to meta shape. The good news is they have a 30 day free trial that you can use, and the software's pretty powerful, and it's also pretty easy to use if we come to their website. Here is and so Meta shape is a pretty versatile piece of software, but if you want, you can get a free trial here by just saying Try it now and pick the platform that you're using. One thing I have noticed between the free software and software that costs a little bit of money is if you're gonna do any really heavy reconstructions with a lot of photos. The only downside to some of the freeware is that the way it manages the memory on your machine when it's doing all the computation sometimes is not quite as well optimized. So if you're not careful with your settings, you can run into some crashes or some issues. Generally, the paid software is a bit better equipped to handle that, but either way, you can still usually get pretty good results. 5. 3D Scan an Object: to start, we're gonna do a reconstruction of an object or a problem. So I encourage you to find something that you want tohave in three D to start. I'm gonna use this little bear model here, and I'm gonna create this photo set up where I'm gonna have a table and I'm gonna set up a old bed sheet. The reason I'm doing this, you don't have to, but the white bed sheet will give us a nice kind of diffused lighting all the way around our object. And it's also going to give us a nice, clean silhouette. It's not essential toe have that, but it just makes it easier for the software to figure out what is your object and what is the background. So you'll see when I take my pictures, imagine a dome over top of your object. I'm going to try to take pictures all the way around that it's important to be cognizant of your coverage because obviously, if you don't take a picture of a certain part of your object, you're not gonna be able to reconstruct that. I usually go for about every 10 degrees or so around the circle, so that It kind of covers every bit of the surface. You don't have to go too crazy because the more pictures you take, the longer it's gonna take for your computer to process them. But I find about every 10 degrees works pretty well. So I've opened up photos, can hear or metta shape, whichever you have. And the first thing I'm gonna do is I'm just gonna come upto workflow here and we're gonna add photos. So I'm gonna pull up here thes air, the bear photos that I took and you can see those guys here and all we're going to do. The first step is we're gonna come toe workflow and align the photos and what this is going to do for now, I think just for an example, I'm gonna show you guys with low accuracy. But obviously, the higher you go, the more points it's gonna look at and the more detailed the results. But if I hit OK, here, it's gonna go ahead and detect those points in the images that we talked about before. So it's gonna look at all the photos, find points, and then, after it's detected, a whole bunch of points. It's going to go through and try to estimate the camera positions. So here's our point cloud. So we have our camera locations. And in fact, if I open up one of these images, if you see this image here, if I click on Ah, this guy here these are the points that it picked out to try to match with other photos, which is pretty neat. So next step, we're gonna go to our workflow, and we're now you could in theory, right away build a mash from this point cloud you have. But we want a bit more detail, so I'm gonna go back and build a dense cloud. So again, I think for now medium is probably okay just for an example. And when I hit OK, what this is going to do, it's gonna go through, and it's just going to using those fixed camera positions give you a whole bunch more points. One thing to be aware off for the dense part is, um, if you look carefully in your scene, there's a bounding box and that's going to find the area or the volume rather that we're going to dig into to create that higher quality point cloud. So just something to be aware of right now this looks pretty good. You're just gonna want to adjust it so it kind of nicely encompasses the area that you want to reconstruct. One thing about doing these kinds of reconstructions. You might want to get a cup of coffee or a drink or something, because sometimes it can take quite a long time. So no, that's finished way. Take a look inside our chunk here. We'll see. We've managed to create a dense cloud. So boom, already. You can see. Here's a lot of detail that almost already looks like a model that's ready to go. And part of the reason this has worked so well is just because of the uniform coverage we got under pretty flat lighting. So if you wanted to, you could use some of the tools up here like here. We have the selection. You could come in and start selecting some of this unwanted stuff and delete it on, and you could do some extra clean up there. But for now, I'm just gonna go ahead and go to the next step, which is building the mesh and what this does. I think for now that all looks fine, you know? You know what? Maybe will go and make a suddenly higher quality mesh. Okay, So what if you remember this generating mesh is doing is it's just taking this dense cloud of points and it's pretty much just connecting the dots. It's a little bit more intelligent than that, but it's gonna take a bunch of points, and it's gonna turn it into a meshed three D model. The mesh generation finished. So this is our dense cloud we were looking at. And now if we open up our model boom, you can see we have a three D model here of our bear. And actually, if you compare it to what we started with, it's already pretty convincing. So the last stage, um, we can already see there some color on this model that it used used the colors and the points that it had in the cloud when it generated this mash. But the last step you want to do is we want to take those original images and project the photos back onto the model, and then we can use that to get a really high resolution texture. So I'm gonna come to my workflow and hit, build texture, and you can customize the size of the texture here and how much it's tiled. Um, because this isn't very big. I'm gonna go with me. Maybe, I don't know, eight tiles off 4000 by 4000 images. So let's go ahead and try that and what this is going to do. It's going to take those images from those camera perspectives, and it's gonna project them from a bunch of different angles. And then it's gonna try to blend those together to get one continuous texture that we can wrap around the whole model So you can see here. It's doing that blending step that I talked about. The texture has now finished, and now when we zoom in, you're gonna see there's a lot more detail on this model here. Wow, look at that. It's almost like a photo, Really. And you can see we've got detail that goes all the way around. How great is that? So the last step here, if you wanted to, you could, um you could, ah, delete some of this unwanted stuff that we have around our bear just to keep the parts that we want. But a to moment I'm just going to export the model as is. And then you can do more cleanup and stuff if you want. So you can see here we could isolate the bear pretty easily by deleting the bits that we don't want. But for now, I'm just gonna leave it as is. So let's make sure we save our work. Always remember to save, but I'm pretty happy with that. Let's just go ahead and export that as a model. So I'm gonna come into file export. You have a few different options here. It depends on what you're using. This four. I'm gonna go with something easy, like an O b j gonna call it bear model. Save that. And then here we have some options about what you want. I'm gonna say include the Vertex colors. But we also want to export this texture here and hit OK, our model is finished. We have our O b j here and you can see the textures that it's also exported, which obviously look a bit ridiculous by themselves. But when you put these all together on top of our model. It looks pretty good. So Mesh Lab is a program I like to use. It's just a free three D viewer. It's really useful. Um, toe, look at models and a lightweight way if you don't want to open them up inside of something else. So here we go. This I don't think the texture is on here right now, but this is what our model looks like. This is what we started with. And if you wanted to model this on the computer, it would probably take you a bit of time. But you can see we very quickly been able to create a three D version. And now this is a prop that you could use inside of a really photo realistic scene. You could use it for your animation, your video game, all sorts of things. So I hope you can see why this is a really, really useful technique. And honestly, it doesn't even take that long. It's pretty. It's pretty easy 6. 3D Scan an Environment: in this next section. Instead of reconstructing something in a controlled environment, we're gonna go outside. I found these logs in my back garden that I think could make an interesting area to three D scan I can imagine. Maybe it's a location I could use in a short animated film about animals in the garden. Or you could also imagine these as like, assets that would be useful to make a video game. When you're taking your pictures outside, there's a couple things you're gonna wanna watch for. The first is if you can try to avoid really harsh direct sunlight, the harsh highlights and shadows air gonna make it harder to do your reconstruction. Generally, if the lighting is a lot flatter and overcast, you're gonna get better results. Another thing I like to do is to lock off the exposure and the focus on my camera. That way I'm just minimizing the amount of variation between my photos when I take them in different angles, like with the object in the studio. When I take my pictures, I'm moving around this area in kind of a dome shaped, taking my pictures facing inwards. Make sure you get the camera up high enough and down low enough because you want to make sure we have enough coverage. So I opened up photo stand here or metta shape, and we're gonna import those photos that we took. So remember, come upto workflow at photos. So I've loaded in these photos here, which you can see I've taken at a bunch of different angles. And then just like we did with the bear, we're gonna come and align the photos first. So when you do this alignment, what it's going to do is remember, it's just going to detect those points in the images and try to match them with other images so that we can train relate to the camera positions Now that we've reconstructed the sparse mash so you can see already we get a pretty good sense of what the shape of this looks like. And these are the points that this tool has found in common between different pictures and kind of used to triangulate the camera positions already. You can see this is looking pretty good. One thing just to keep an eye out for you'll see, there's this red kind of square, the bottom here. This is defining the bounding box, and that's going to find the area or the volume rather that we're going to dig into to create that higher quality point cloud. Just something to be aware of right now this looks pretty good, but if this ever was, say, too big or too small, you're just gonna want to adjust it so it kind of nicely encompasses the area that you want to reconstruct. So when you build a dense cloud on, I think mediums okay, for now, it's gonna go through and it's just going to using those fixed camera positions give you a whole bunch more points. So now that that is finished, you can see here we have a whole bunch of points in here that are giving us an idea of what this looks like in three D. So already this almost looks like a finished three D model. And because it's using the colors of the pixels for those points, it already almost looks like it's textured. So this is looking pretty great. I'm just gonna go ahead and go to the next step, which is building the mesh and what that's really going to do is it's just going to go through and connect those dots and create an actual three D model. So let's come into our workflow here and build a mesh. I think a lower quality is gonna be fine for this guy because we only sort of care about the rough shapes. So I'm gonna go ahead and build that, uh, build that model. Okay? And now that that's finished, we can see here. This is what our three D model looks like. So now we have an actual mash. Um, you can see if you zoom up close here. The texture isn't on there yet, so it's just kind of blobs of color, but it's already giving us a sense of the shape. And if I come over here to the unshaded mode, you can see this is what our model actually looks like. So for the last step, all I'm gonna do is I'm gonna go to my work flow on. We're gonna build a texture. So what this is going to do is it's going to project those those images that we started with back onto the three D model, and it will give it a texture that has a lot more detail. So I'm going to That looks pretty good. I'm gonna go with a 8000 texture. And this says we're gonna tile that to be different times. So we're gonna get eight times this resolution squared, which is quite a lot. So I think that looks pretty good. So let's go ahead and run this guy. So now that the texture is done, if you remember, this is what we had before. But if you click on this icon here, you'll be able to see her model with the texture projected onto it. And there it is. There's are finished three D model and you can see it's not perfect, but it's actually got quite a lot of detail in it. I'm pretty impressed with this, and I see in the background here there's a few kind of unwanted bits, and if you remember, we can get rid of those by just selecting bits and deleting them. So I'm gonna use the selection tool here just to kind of select a few bits here, and then I'm gonna hit to the remove button and there it is. So obviously you can go through here and do a lot more clean up. Um, and you could do a higher quality version as you see fit if you're willing to wait for it. But I hope that shows you that pretty quickly we can get a model that looks pretty good. So I'm gonna go ahead and export this export, uh, model, and it's up to you. What? You save it as depending on what you need. I think sticking with a good old um O B J is an easy place to start worn FB X. So I'm gonna go ahead and export this guy, and you have a few options here about the textures and stuff like that, but I think that looks pretty good. You could maybe imagine if you were making say, like a short animated film about critters in the garden or something like that. You could use this as your set, which would save you a huge amount of time in building that yourself by hand. And even if you go in afterwards and do some clean up and touch ups, it gives you something to work with really quickly. So obviously there's more stuff you can do here to try to refine these models and adjust the texture, ring and stuff like that. But really, my goal here is just to make sure you guys can very quickly get up and running on start making three D models of your own. 7. Conclusion: thank you very much for watching my class. I hope you learned something. And, uh, I hope you can see just how surprisingly easy it is to take your cell phone and use it along with some software that you can get pretty easily to create some surprisingly realistic three D models of stuff that would take you a lot of time to make by hand. I really encourage you to try to reconstruct some models of your own. I hope you find this really useful because it's amazing what it can dio. I'm excited to see what kind of stuff you guys come up with, and I'll see you next time.