Photogrammetry Apps Comparison:
123D Catch vs. Trnio vs. Scann3D
3D Scanners are getting more mobile as we speak. The Structure Scanner I reviewed recently can be attached to an iPhone 6 with a special case. And Matter and Form has developed a 3D Scanning add-on for Android and iOS smartphones—the Bevel—that will be available soon. And this summer, you can buy the first phone (allright, phablet) made by Lenovo with Google’s depth sensing technology—Project Tango—integrated.
It makes sense: we’re already capturing the world around us in 2D with our smartphone cameras while our dedicated devices lay in a drawer somewhere. So it’s natural that we want to capture 3D wherever and whenever we want and share it directly online.
But while depth sensors are indeed getting small enough to be pocketable soon, pure-software photogrammetry solutions that can generate 3D objects from regular 2D photos are getting smarter and faster, too. So I started asking myself:
do consumers actually need dedicated 3D capturing hardware on their phones if software can do the trick?
To test the current state of software-only mobile 3D scanning, I tested four different smartphone photogrammetry apps: 123D Catch, Trnio, Scann3D &
Seene on the same object under the same circumstances.
If you’re a regular reader, you know that I already reviewed this app a while ago, but lets recap (pun intended): 123D Catch let’s you take up to 70 photos of your object from different angles and assists you in doing this with a 360° progress indicator. After this, your photos are downscaled to 3 megapixels and uploaded to Autodesk’s cloud-based photogrammetry engine, the same one they use for their professional software suites.
Thanks to the downscaling, the uploading goes reasonably fast, of course depending on your internet connection and the amount of photos you took. Even uploading 70 3 megapixel photos might not be a good idea on a mobile dataplan, so you’ll probably want to wait before you have WiFi. You can only do one capture at a time that way, because the photos aren’t saved (I just discovered it does only give you that option if processing fails) and you can’t load existing ones from your smartphone. Cloud-processing speed depends on how long your projects gets queued, but waiting half an hour is certainly something to expect. Unfortunately, the app doesn’t notify you when it’s done.
For the sake of good comparison, I redid the capture (with an iPhone 5S this time) of Teddy from my original post. Here’s the workflow of 123D Catch in Capture, Processing and Preview stages:
There’s a 3D Preview, but no way to edit the result. This way you can only share the result with all background clutter included. From the app itself, you can Share it to the Autodesk’s own 123D Gallery where other 123D users can like and comment on your work. You can also share the link with any social app you have installed which points to a page like this. It’s a shame that you have to click the (ugly) “3D View” button below the image to see the capture in actual interactive 3D. And even if you do, finding my Teddy Bear is like Where’s Waldo?—it’s even upside down! Far from intuitive.
You can’t export the model file from the app itself, but logging into 123dapp.com does allow downloading it in various formats like .OBJ and .STL.
123D Catch is available for iOS, Windows Phone/Tablet and Android. Unfortunately it’s incompatible my Google’s Nexus 5X because the app still uses a depreciated Android Camera API.
Even though the 8 megapixels of the iPhone 5S camera got downscaled to 3 MP before processing, the result is impressive. No less than 651.6k faces—from only 36 photos! Since the app hasn’t been updated since my original review, Audodesk still didn’t implement any smart automatic cropping or a manual clean-up feature, so there’s stray geometry everywhere. This should really be cleaned up before being usable.
That being said, the geometry is simply stunning. If you turn on the MatCap Rendering setting in the Sketchfab embed, you’ll notice that even without the textures there’s a lot of detail. The different knitting patterns of the shirt and body of the bear are clearly visible and even the wooden skewer is captured.
Part of the Hacker Unit accelerator program, the developers of this app have taken another approach to capturing the photos required to generate a 3D model. Instead of taking separate pictures, the app will initiate a burst mode that automatically takes a maximum of 70 photos while your path around the object is being tracked. This greatly speeds up the capturing process.
Unlike 123D Catch, Trnio does allow loading existing photos from your iPhone’s Camera Roll. You can even send an upload link from the app to your email address. This creates a new project in the app and allows you to upload photos to that project from your computer. That feature is still Beta though and only accepts photos (80 max) in a .TAR file, which is a kind of .ZIP-file for servers that requires special compression software.
After capturing, downscaled versions of your photos are uploaded to the cloud and processed. In my case processing was a little faster than 123D Catch, but that might have been because the queue of this relatively unknown app is less crowded. When the result is downloaded, you’ll get a notification (Yes!) and can view the Preview. The app features an ingenious cropping feature that allows you to select which parts you want to keep by making a selection from the top and side view of the model. That doesn’t give you total cropping freedom, but works pretty fast. Unfortunately, the model has to be re-processed in the cloud to be cropped.
If you like the model you can also choose to upload the full resolution photos for better textures. The model has to be reprocessed for this, but the result is worth it.
The Trnio app itself is also a kind of social network. It allows others to follow you and like your work. You can make a profile similar to Instagram, but somehow my own captures don’t show up on my profile page. Like Instagram in its early days, you can’t share a link to your capture on other social networks directly. You can, however, export your capture to the online 3D showcasing website Sketchfab, which is no surprise, since Sketchfab’s CEO Alban Denoyel is one of the mentors at Hacker Unit. This procedure does require you to visit Sketchfab in your browser and copy & paste the API token to Trnio, but only the first time.
You can supposedly also send a temporary download link of the model to yourself through email, but the link I received didn’t work at the time of writing.
Trnio is currently iPhone-only, but an Android version can be expected later this year.
If you set the rendering of this embed to MatCap, you can see the difference in texture on the arms and the edges of the shirt. Teddy’s eyes are also somewhat visible without the textures. The model has 159.6k faces, which is actually quite good.
This Android-only app developed by SmartMobileVision doesn’t feature any social features yet, or even an account registration system. But after logging in with a temporary username, it does have features that make it different than the apps above. The most important one is that it doesn’t do cloud computing but instead does all calculations locally on your phone. While the speed of this greatly depends on your phone’s processor—and naturally drains the battery—it worked surprisingly fast on my Nexus 5X. You can do other things on your phone while waiting, but that impacts the waiting time. The upside is that you don’t have have to upload and download anything while making captures on the go.
Both the 3D Model and the Photos you took are stored locally under “My Models”. You can also import image sets here, but they have to be numbered a certain way (name001, name002) and be a complete sequence. From here you can (re)start the processing of an image set.
One other thing that sets Scann3D apart is that it offers fine control over the capturing and processing quality. There are Low, Medium, High and very High Presets, but you can also enter Advanced Mode to enter custom settings. The Image Size is capped at 1920 x 1080 pixels, anything larger is probably unrealistic for phone-based local processing.
You can export the 3D model as .STL, .PLY or .OBJ file directly from the app. It saves it to a dedicated SCANN3D folder on the local storage of your phone. This way you can either send it to yourself through email or upload it to your favorite cloud storage, like Dropbox. Just remember that on newer versions of Android, you have to turn on “view local storage” from the 3-dot menu in the left corner of the file browser. Then the Local Store will appear in the drawer on the left, letting you browse to the correct folder. It would be nice to see native Android sharing (to Dropbox, Google Drive, etc) from the app.
Scann3D can also share directly to Sketchfab and is one of the few apps I know that doesn’t require users to manually copy & paste the API token, but instead login to Sketchfab directly from the app itself to authenticate.
It’s good to know that the app—which is currently in Beta—is being developed very actively. Even within my reviewing period, I got an update with a new viewer to try out! The geometry will probably improve with each update as well.
The High version has 73.7k faces and the Very High 84.7k. That makes the polycount about 10 times lower than 123D Catch and half of the that from Trnio, which is visible in the details of the mesh. As you can see in MatCap rendering mode there are no actual details in the geometry. As you might expect, the Very High version’s textures are a lot sharper.
Seene is the most polished app of all four and also beats them when it comes to social features. It’s approach to 3D capturing is also the easiest and fastest. But that’s because it’s also the least 3D-ish of them all: Instead of 360 degrees of capturing freedom, Seenes—as they’re called—are limited to about 10-20 degrees. That greatly reduces it’s usability, but that’s not the point because Seenes are more like 3D snapshots than complete captures. This approach, combined with a wide selection of filters, makes it the best attempt to “Instagram in 3D” I’ve seen so far. Just like the popular photo app, Seene let’s you create a profile that can also be accessed from the web—like this awesome one—so you can share it with people that are not members (probably everyone you know).
It makes sense that the developer has recently been acquired by Snapchat, the fast-growing social network that has a lot of 3D tricks of its own.
While it’s not sure what will happen with the Seene app itself The Seene app has been discontinued as of July 2016—Snapchat did supposedly buy the company behind it because it has wide array of object tracking technologies that it might implement in it’s own app.
These tracking technologies are partially what makes Seene itself so fast: point your camera at something, move a bit for a single second and you’re done capturing. Plus the locally processed result will appear almost instantly, ready to share. This can be done within Seene itself, which I discovered has a surprising amount of active users. But you can also share a link to your Seene on other networks. It even includes a still frame when you do so. From the shared website, you can also embed interactive Seenes directly on your website, without the use of another service such as Sketchfab. There’s no way to export the 3D file anyway and that wouldn’t make sense to do so.
You can load photos taken earlier, but only if they where captured with the Lens Blur feature in Google’s Camera app. This basically saves a depth map into the JPG file. (If you want to know how this technology—called Structure-from-Motion (SfM)— works, read this).
The 3D snapshot technology works better with more complex scenes with lots of tracking information. I’ve some better examples on my Seene profile. But no matter how hard you try, you will almost always get stretched textures on places that the camera didn’t see because of the limited difference in capturing angles.
There’s no real winner here. If you’re looking for the best possible result, 123D Catch is the best app and it’s available on every mobile platform. It simply allows you to tap into Autodesk’s industrial grade photogrammetry algorithms—for free! But it really falls short when it comes to smart cropping and sharing requires manually downloading the model from a desktop computer. Without the cropping feature 123D’s own social gallery is a mess.
Both cropping and sharing can all be done from the Trnio app directly, but the former requires re-processing and up/downloading of the model. This time is partially countered by the faster automatic capturing. It has social features, but they’re in-app only, so you can’t share links on other networks.
Scann3D has the best interface design in my opinion and differs from the other by being totally offline—both for storage and computing. This has the advantage of not having to upload photosets to the cloud when you’re on a mobile data connection, but the downside of local computing is that it drains your smartphone’s battery quickly. It also shows the limits of the amount of geometric detail that local processing can achieve at this moment, but this could change when the algorithms improve over time.
Seene isn’t comparable to the above, but it shows that 3D can truly add something for a social network when capturing is almost instant and sharing is easy and flexible. More specifically, it shows what’s realistically usable today with photogrammetry on a consumer level, with current smartphone hardware and internet speeds.
To round it all up it’s clear to me that software-only 3D capturing has one very big problem: speed! Many people have asked my lately why they should invest in a 3D scanner if they can capture the same objects through photogrammetry for free and the answer is also: speed (plus interactively seeing what you’re doing). That might not be important for creating 3D assets in an environment where time is no issue, like a hobby or even some professional cases. But all apps above are aimed at making 3D scanning accessible for people who live in a world that’s increasingly realtime. We expect that the camera’s on our phones can be turned on instantly with a double button click so we can shoot fast, share fast and move on.
This speed does come at a price though: here’s the same Teddy Bear 3D scanned for my Structure Sensor Review with the stock app that doesn’t require cloud-processing. It has just 16.2k faces—less than any of the free photogrammetry apps above!
The Structure Sensor is portable, but a tad too large to call it pocketable. And while Google Tango is indeed soon available in a Lenovo phone, that technology is not necessarily meant for 3D scanning. It’s purpose it’s positional tracking indoors and the applications lie more in measuring, enironment mapping and Augmented Reality. Tango can be combined with Intel RealSense, though, which is capable of 3D Scanning, but for smartphones this concept is still in the development kit phase. I’m really curious to the Bevel smartphone add-on from Matter and Form. That seems pocketable, but will it’s captures be good enough? Hoping to get my hands on one soon.
I hope you enjoyed reading this post. If you have any questions, please drop a line in the comments or send me a Tweet. And if you think this article is useful for your friends and followers, I’d appreciate it if you share it by using the buttons below.