Back to Blog

Getting Started with Photogrammetry Using Your Cell Phone

By Patrick Letourneau
Cinema 4DRedshift
Sometimes the best way to create a new 3D asset is to capture it from the real world. Welcome to Photogrammetry!
How do you take an object from the real world and bring it into Cinema 4D? You could spend a few hours modeling it yourself...or you could use your cell phone, some free software, and the power of photogrammetry.
Hi, I’m Patrick Letourneau: 3D artist, photogrammetrist, and secret crime fighter. You’ve probably heard the term Photogrammetry before, but maybe thought it was a bit too advanced or complicated to try yourself. I’m here to show you the technique for capturing incredible 3D scans of the world using the tools you have at your fingertips.
In this tutorial, you’ll learn:
  • How to set up objects for capture and import them into the software
  • How to clean up and simplify your model
  • Texturing and baking normal maps
  • How to export the model into Cinema 4D and Redshift

Download the free project files

Before we begin, don’t forget to grab the project files in the description below so you can follow along. Now let’s get started!

Be warned! This is a 5GB pack of free files and textures.

Download Now

Getting Started in Photogrammetry Using Your Cell Phone

What is Photogrammetry?

Photogrammetry is the science of making measurements from photographs. Using multiple input images, software is able to infer super accurate 3-dimensional models that you can put to use. This process may seem complex, but it can be much faster than modeling new assets from scratch. Better yet, you don’t need expensive equipment and complicated software to get started...just your cell phone and some supplies from around the house. 

How do you get started in photogrammetry?

In order to get started, you just need a camera, a computer, and an object.
Okay, that may have been a little too simple. The bare minimum you need is a camera and a computer, but you will obviously get better results with more advanced equipment. What I aim to show you today is what you can achieve with what you already own.
You don't need any prior knowledge or technical skill to get started, but a few things will certainly help. You need to understand basic photography, including lightning control. You'll also want to make sure your object is set up in a way that allows you to capture every angle.
Start with something small, like a shoe, before you graduate to larger captures.

Lighting matters most

When you capture an object using images, any lighting will be hard baked into the final asset. That's why it's important to have consistent, neutral lighting. Shooting outside on an overcast day is good, using a soft box is better, and building a rig with cross-polarized ring lights is best.
Watch out for harsh lighting, shadows, or direct sunlight.

Camera control is important

If you're using a DSLR, you'll have access to all the controls you need to capture consistent images at a high-resolution. If you're using a cell phone, you'll likely need to download a photography application that enables exposure and aperture control, raw images, and consistent brightness.

Take pictures from every angle

Once your object is set up (using a tripod, rotating base, or whatever you can MacGyver together), it's time to take some photos. Make sure you make several passes at low, mid, and high angles to capture every side of the object.
Depending on the camera you use, you can select how each image will be captured. Using RAW file formats is best, as that contains minimally-processed data from the image sensor and produces the best results.

Prepare images for transfer

Before you transfer the images, prepare and organize them for an easier export and better results. If you have to use JPG, lock the white balance and lift the shadows to clean things up. If you're a DSLR user, you can check the color to make sure each photo matches, though this step can be skipped if that is too advanced for you.
For the best asset, I recommend TIFF, but that can be memory intensive and much slower. Since the software I recommend is a "pay per input" model, this can also make it a more expensive endeavor.

What software to use for Photogrammetry?

Once you have your photos ready to go, it's time to upload them into the rendering software. I recommend Reality Capture, a pay-as-you-go program that provides some great results.
To get started, download the program and open a new project. Drag and drop your images into the project and you'll see them in the bin on the left side of the screen.
You could just hit START and let the program work out the rest, but let's go through a little slower. Begin by hitting ALIGN. This tells the program to sort through the images and asses the camera's position for each shot, which should create a more accurate 3D model.
Each cone represents a shot from the camera. This can be useful in locating gaps in coverage, but for our purposes this is looking really good.
Now that we've checked our alignment, quickly check that all the images had their positions solved (look for flags on files in the bin). Then we can calculate the high resolution model.

How to touch up your photogrammetry scan

Now that we have our model calculated, let's have a look.
Not too shabby for a cell phone! In fact, this looks pretty good. I can see some jagged edges, a bit of noise in the smoother areas, and we'll obviously need to get rid of the toilet paper roll, but there is no mistaking what this object is.
It should be said that a higher-quality capture (using a DSLR) will provide a more accurate scan and an easier finishing process. Right now, we have two issues: This object is made of 15 million triangles, and it needs some clean up. So let's get started.
Go to the Reconstruction Tab and grab the Lasso.
Drag the lasso around the stand and it will highlight. Then go to Tools > Filter Selection.
Now we have a shoe without a toiler paper roll underneath, which is great (unless you were trying to create a shoe with a toilet paper roll underneath, in which case you just messed up big time).
Next we're going to hit Close Holes in the Tools menu, which will pop up a menu on the left side to fine tune the results.
Click Close Holes and the program will auto-fill the area where the toilet paper roll used to be. Now we have a nice, plain shoe just waiting to be finished. Next, we need to simplify the model (15 million polygons is a bit much).
Go back to the Tools menu and select the Simplify Tool. Then, we just hit Simplify.
Now we have a model that looks great, if not quite as detailed as before. You'll notice the laces seem less defined, and there are harsh edges. We need to fix our low poly model before applying our high-poly texture.
Go back to Tools and select Smoothing Tool. In the pop-up window, increase the iterations to 5 for a cleaner pass. Then let the program go to work.
Now we have a smooth, if somewhat melted version, but this will be a much better target for our projected textures. This is also a better target for baking out our normal maps.

How to texture a photogrammetry scan

Now it's time to texture our model. First, we'll make sure we select the low-poly version. It's a lot easier to work with 250,000 polygons than 15 million. Then, we just hit Texture. No need to mess with the settings; the program can handle it from here.
Yowza, that's a fine looking shoe
You can see the benefit of shooting on an overcast day, as we have relatively few baked-in shadows (there's a little underneath the laces, but that's unavoidable without a more professional lighting approach). Overall, this looks pretty dang good.
Most of what we're seeing is ambient occlusion that can be processed out so we can light from scratch. Now that the textures are done, it's time to bake on the normal maps from our high-poly model.

How to bake normal maps to a photogrammetry scan

First thing we'll do is deselect the Smoothing Tool and click on the Texture Reprojection tool. Our Source Model will be the high-poly version, and the Result will be the low-poly version. Then we hit Reproject.
With the Normal Map baked, we get a handy diagnostic view that allows us to diagnose some of the noise. This won't be as much of an issue with higher-quality captures.
Now we can export the model and bring it into our software of choice. In this case, we're headed to Cinema 4D.

How to export your photogrammetry model to Cinema 4D

When exporting from Reality Capture is as simple as a few clicks. Be sure you're selecting all the layers, including the textures. A JPG is fine for diffused textures, but textures and displacement should be as uncompressed as possible.
Once Reality Capture exports the files, you can pull them into the rendering engine of your choice. Check out what we do in Redshift in the video above!

It's a Photo Finish

Those are the basics of Photogrammetry. It’s pretty impressive what we were able to achieve with just a cell phone, a tripod, and a toilet paper roll—but the difference in quality when moving to a professional set up is dramatic. If you want to try some out, be sure to share them on social with the hashtag #nogoodphotogrammetrypuns

Cinema 4D Ascent

If you want to learn how to get the most out of your 3D assets, check out Cinema 4D Ascent. Over the course of 12 weeks, you’ll go from beginner to intermediate-level 3D artist that’s fluent in Cinema 4D and familiar with other 3D tools.