March 17, 2023
Creating 3D assets from real world objects is easier than you might think. With practice, you’ll be able to achieve unmatched realism!
Photogrammetry ia a great way to grab amazing details out of real world objects. In this guide, we’ll go through the type of gear you’ll need, how to properly go about scanning an object and lots more little tidbits. Let’s dive in!
As complex as the math behind photogrammetry might be, understanding how it works is actually really simple! Let’s say, we want to scan a statue. What we have to do is go around and start taking pictures from all angles. Top, bottom, sides, every angle you can imagine. Then, through the use of specialised software, those photos are analysed to find common points and features in each image. Based on the camera’s lens and sensor the software can calculate the distance, depth, and position of each of those points. If we did a good job taking overlapping pictures, the computer will have enough information to create a 3D model.
It’s kind of like piecing together a puzzle. The result is an incredibly detailed model that shares all the qualities of the real thing. The imperfections, the dents, the faded material, it’s all there!
So now that we know how photogrammetry works, what do we need to start scanning? Let’s divide things to smaller chunks.
Photogrammetry is one of those areas where you could spend a lot of money, but it doesn’t have to be that way. When first starting out you could just invest on the bare minimum gear. Just a camera and a lens. You can always use natural lighting to scan the objects. The results won’t be perfect but it’s better to see first if photogrammetry is something you want to pursue. rather than investing a whole lot of money only to find out you don’t really like working with photogrammetry!
For a full blown setup we will need:
Any camera of the past few years will do. Nowadays the majority of the cameras in the market are extremely capable. Even the camera on our phone will do. As long as we can manually control aperture, shutter speed, ISO and white balance we’re good to go. I personally use a GH5 but there are better cameras out there. The bigger the sensor the more detail we can capture but you don’t have to go crazy. You don’t have to spend thousands of dollars just for a body!
These two cameras can work great for photogrammetry. The Canon R10 is a lower budget model while the Sony A7 IV is a more high end one. Both will do a great job.
Canon EOS R10 | 24.2MP | $879
When it comes to lenses, any lens in the range of 24mm to 80mm will do. It really depends on what focal length you already have available and what the subject matter is. A 24mm lens is great for capturing outdoor environments or bigger structures. On the other hand a 60mm lens is great for small objects. Whatever the case may be though, it’s important to use a lens with low distortion. That way you can minimize any potential image deformities that could affect the accuracy of the final 3D model.
When it comes to lighting, it’s best to use consistent, diffuse light to reduce extreme shadows and highlights. Both of them can affect the accuracy of the 3D model and the texture map. For outdoor scanning prefer overcast days. That’s when the light is soft and evenly distributed. When shooting indoors, you can use multiple artificial light sources, such as softboxes or diffused LED panels. That type of setup will create a uniform and diffuse light. Avoid harsh or directional lighting that can create strong shadows or cause reflections on the object. I personally use 2 Falconeyes RX18-T. It’s a cheap and relatively bright lightsource. And most importantly since it’s LED it stays cool when I capture things I don’t have to literally sweat from all the heat!
This is also incredibly important. At the beginning you can skip using cross-polarised light altogether, but at some point, it's a good idea to switch to that.
So what does cross-polarisation do exactly? It mitigates the impact of surface reflections and highlights. When both the light source and the lens use polarisation (cross polarised) we can actually capture the exact diffuse properties of the object.
Linear Polarised film for the light
Gobe polarise filter for the camera
By placing an object on a turntable and rotating it incrementally, we can capture a series of photos from different angles effortlessly. This controlled rotation ensures consistent coverage, allowing the photogrammetry software to precisely analyse the images and create accurate 3D reconstructions. Turntables also save time and effort by automating the rotation and picture taking process, eliminating the need to manually rotating the object for each shot.
When placed as a backdrop and/or beneath the object, a black or white sheet acts as a high-contrast and less busy background. That makes it easier for the software to identify the edges of the object, reducing background distractions. and not lose focus on other elements that might otherwise be in the background. On top of that, a white sheet can help with the lighting of the object since it will reflect back to the object. The easier we make it for the software the faster and more accurate our scans will be!
Tripods and stands are necessary for your camera and lights. You don’t have to spend a fortune on them -even though you can – just buy some that have decent enough features. For example the tripod I’m using has a rotating head which makes camera framing a whole lot easier.
Lightroom
Lightroom allows us to fine-tune and optimize our photographs before feeding them into the photogrammetry software. No heavy editing is needed here. We just need to balance things out. If a picture is darker than the others we need to lift it up so it matches the rest. We can also do other small improvements like brightening up the shadows and bringing down the highlights.
If you’re in a controlled environment you won’t have to spend so much time on this part. The great thing with Lightroom is the batch process feature. We can easily copy an adjustment from one image and paste it to multiple others. And of course we have an extremely easy way to export all these images in a matter of seconds. Something worth noting here: Avoid using any noise reduction operations. These can mess up the surface of the object and will make the photogrammetry process difficult or in some cases impossible!
With our edited pictures at hand it’s time to produce the 3D model. To do that we’ll use a specialised piece of software. There are a lot of different for this. If you’re on a PC you have a lot more options and if you’re on a Mac you’re a bit more limited but you can still get amazing results whatever your platform might be, so no need to worry about that!
Metashape (Mac & PC):
A really good solution that works on both platforms. The interface might look scary and things can work in weird ways but the results are really good. Not the fastest option when it comes to the final solve but with some tweaking we can get a good compromise between speed and quality. The regular version is more than fine. No need to go with the Pro especially when you’re first starting out.
Reality Capture (PC only)
Reality capture is the software to use, if you own a PC. It stands out for its remarkable processing speed and ability to handle large datasets, making it the standard for people working on photogrammetry. They have a really weird payment system where each model costs a certain amount of credits. The amount you pay depends on the number of images and megapixels. Since I’m a Mac user, I haven’t used the application, but everyone swears by it so I would say it’s a pretty safe bet
Photocatch (Mac)
This might look like a strange addition since the app looks like a toy, but it’s actually a really good solution. The reason it’s so good is because it’s using Apple’s photogrammetry API. And let me tell you the results are incredible. Because it’s optimised for the Mac it’s using every single resource available so the solving happens amazingly fast. It’s the program I personally use. Sometimes I will use Metashape but the majority of my scans happen on Photocatch.
There are a million other photogrammetry applications. For example Adobe’s Substance Sampler, but I would say stick to the suggestions above. Substance Sampler for example uses different algorithms for the Mac and PC version. For the Mac they uses Apple’s SDK so no reason to pay for a subscription when you can use a completely free solution like Photocatch
Zbrush
After the model is created we need a way to clean it up and prepare it for final output. I use Zbrush for pretty much all aspects of the production.
1) Cleaning up the high res mesh 2) Preparing the low res version 3) Creating the final textures
It’s basically a one stop shop and a great modeling/sculpting overall.
Delighter
This little utility does one single thing. It makes the darker parts brighter and the light darker. As simple as that may sound it’s an important step when trying to even out the lighting of the texture. Remember we need a nice even lighting in order to get a good clean diffuse map.
Cinema 4D or other 3D applications
Once we have the final low res mesh and textures we need a way to put everything together. Create a material for the renderer of our choice etc. This is where Maya, Cinema 4D and other applications of that sort come in to play. I personally use Cinema 4D but you can also easily use Blender which is a completely free solution.
I have several videos about all parts of the process so I will leave a couple of them here. The must watch are the following two. Cleaning up a scan with Zbrush and understanding the steps when delivering a final piece of asset.
For more photogrammetry content make sure to visit my channel on youtube.
You can find the photogrammetry playlist here