Today, we’re interviewing Alex Schiff, product manager for Canvas, a new 3D scanning app that makes it fast and simple to capture a 3D model of a home… for use in SketchUp, of course!
So, what is Canvas?
Canvas allows you to create a 3D model of a home in minutes — all on an iPad. You can view, share, or measure your model on-device, or use our optional Scan To CAD service to convert it into a design-ready, professional grade CAD file. It’s powered by the Structure Sensor, a 3D sensor for mobile devices made by the same company (Occipital).
Put simply, Canvas is mobile 3D scanning that “just works.” Just enter a space with your iPad and Structure Sensor, hit “Scan,” and you’ll see a 3D model of the scene build itself on-screen as you move around. We can help you skip right to this for every project:
Why did Occipital build Canvas?
Occipital has been wanting to bring something like Canvas into existence for years, even before I joined the team to build out a focus area we call The Augmented Home. If you look back at our Structure Sensor Kickstarter video, you’ll see many of the uses we had in mind actually revolved around the home. It has just always seemed extremely obvious that having a 3D model of your home is going to play an important role in the future: to quickly capture measurements, visualize changes before you make them, or simply share with other people for any number of purposes (service providers, family, etc).
Even though we knew that the payoff would be there, a lot of puzzle pieces had to come together just right to make Canvas possible. First, we had to really level up our software to go from scanning objects (the main early use of the Structure Sensor) to entire spaces. For the tech-inclined readers, I recommend reading up on “Simultaneous Localization And Mapping” (SLAM) — it’s a really fascinating research area within computer vision, and it’s basically what we had to make “just work” on an iPad. Second, even after we got things technically working, we had to iterate on the UX over and over again until an average user could get up and running without tons of hands-on guidance. Finally, even after all that work on the underlying scanning experience, we had to figure out how to bridge the gap to existing industry tools like SketchUp.
Spark notes version: this stuff is really hard, and it took a lot of time and manpower to get to where we are today. In fact, Canvas was the kind of product we hoped people would build on our own platform, but eventually we just decided to take matters into our own hands. We wanted to bring the world closer to the future we all knew was coming — one where 3D is the default basis for making decisions about your home.
Who has been using Canvas so far?
Canvas is made for anyone who wants to capture a 3D model of residential spaces — be it their own or someone else’s.
Our primary customers tend to be design-oriented home improvement professionals (like interior designers, design-build firms, and architects), but we have customers from pretty much every nook and cranny of the “home” ecosystem. In fact, some of our highest-volume customers are retailers using Canvas to power in-home visualization of their products, not selling services.
Outside of home services, if someone’s job relates to understanding physical space, I’ve probably seen someone try to apply Canvas to that context. Crime scene reconstruction, facilities management, events, film, you name it. I even used it to scan “nuraghes” (village and tower ruins from the prehistoric Nuragic people) on a recent trip to Sardinia!
You say Canvas can measure an entire room in only a couple minutes? How does that work?
Sounds like sci-fi, doesn’t it? Watching space become digitized before your eyes certainly can feel that way!
First, you’ll need an iPad, Structure Sensor, and optionally, our Wide Vision Lens. The Structure Sensor allows your iPad to measure depth information, which is why you can get true, scale-accurate measurements from your Canvas-captured 3D models. The Wide Vision Lens helps Canvas maintain the best possible tracking on the scene, which enhances the quality of your results. You can buy both the Structure Sensor and Wide Vision Lens as a bundle on canvas.io for $399 (vs. $418 if you buy them separately).
Once you have the sensor attached, calibrated, and the Canvas app installed, you’re ready to start scanning! Just hit “Scan”, and you’ll see a white overlay start to appear over what the camera sees on your iPad screen — kind of like you’re spray-painting. This white overlay is actually giving you real-time feedback about what you’ve captured and what you haven’t, so you just move in a loop around the room until you’ve captured all the details that matter. It looks like this:
Once I’ve scanned a room with Canvas, what can I do with that 3D model on my iPad?
The model you see on device is called the 3D mesh. This is the reconstruction of the space created by Canvas and your Structure Sensor, and you can use it immediately to get measurements, navigate to different views (including a top-down view), or share via email. This is what a 3D mesh looks like:
Working with a 3D mesh is usually new for the home improvement crowd, and touching the exact point you mean to measure can sometimes be a challenge. To help, we try to be smart about what measurements we think you’re trying to capture and automatically “snap” to the correct points.
Plane-to-plane is the only feature like this right now, but we plan to release more in the future to speed up common workflows. If people reading this blog have specific requests, I’d love to hear in the comments!
Of note: this 3D mesh is 100% free to use, and you can export it as a .obj file via email to use however you want. Unfortunately, you need somewhat specialized tools to actually measure or otherwise work with this kind of data, and they’re usually not the same ones used by the architecture and design community. You can sort of import a scan into SketchUp using special plugins like Fluidimporter Lite, but it really slows down the program, and the model doesn’t have any of the semantic understanding you need for design (separate faces, layers, groups, etc).
Okay, so how does this data find its way into SketchUp? In the past, this has been a major point of friction between scanning and useful 3D design.
You’re 100% right about that friction, and that’s exactly why we built Scan To CAD — it’s what really makes things “just work” for SketchUp users. Some readers might be familiar with 3D scanning from expensive LiDAR systems, and I’m sure they still have nightmares about the hours spent “tracing” the point clouds generated to form a usable CAD model. And that’s assuming you can even figure out how to get the massive amounts of data generated from the site to a computer that might be halfway across the country.
Scan To CAD is a turnkey service we offer to convert 3D models captured with Canvas (what you see above) into editable, professional-grade, design-ready CAD files. You just select which scans you want to convert, pay $29/scan (one scan covering one typical residential room), and two days later you get an email with those scans converted into .skp, .dae, and .dwg format, as well as an enhanced, colorized version of your scan. A typical SketchUp result might look like this:
You’re of course welcome to do the conversion manually, and we have plenty of customers that do this. However, by using our pipeline, you’re able to parallelize your work and get started on the more fulfilling parts of the project (design, materials selection, planning, etc) while you wait for your files. You get to skip ahead to the fun part!
How much time is this really going to save a typical SketchUp user, vs. the old fashioned way?
We actually ran a handful of pilots with SketchUp users right before we launched in November 2016, and they reported average time savings of about 7-10 hours per project with Canvas. That can add up to a lot of time (and therefore money) saved over the course of a year.
"SketchUp users.... reported average time savings of about 7-10 hours per project with Canvas.”
Moreover, we really focused a lot on making things “just work” for SketchUp specifically. Very early in our prototyping, we sent a SketchUp user a 3D shell of a room they had scanned and said, “Is this what you need?” She responded, “Um, I can’t hit explode!” I had never even heard that phrase before, but it revealed that just producing a 3D block of the outer dimensions — something that computer vision can feasibly do automatically — isn’t enough.
The SketchUp models you receive from us are organized, properly layered, and, as simple as this sounds, each face is still independently operable. This “last mile” of CAD is really hard to automate, which is why we have a human in the loop to ensure the results are always professional-grade and design-ready. We decided very early that we’d rather have a solution that works 99% of the time and costs a little money vs. a free solution that only works 70% of the time and can’t be used in true professional workflows.
7-10 hours is a pretty massive difference. Where is that time going? Are you finding that people are changing their workflows as a result?
I tend to see a few different types of workflow changes happen pretty quickly after Canvas takes root, besides simply saving time:
- Capturing more data on-site to upsell larger projects. Where you once might have spent 45 minutes getting the measurements needed to put in a new floor, you can now capture a 3D model of the entire house in the same time. That means that when you come back to the client with an estimate for the flooring, you can also provide a few concept ideas and bids for other, larger projects, like opening up the kitchen into an open floor plan.
- Jumping into 3D earlier in the project lifecycle. Currently, 3D often doesn’t enter a residential project until the very end. It’s often a last check on the design plan to get the client excited and ready to write a check. Because Canvas drives down the cost (in time and in $) to get the as-built conditions into 3D, we see people integrating 3D design into their process earlier — either to start iterating on design concepts faster, or simply to win the bid with a higher-fidelity pitch.
- Expanding 3D to all clients. When given the choice, every client wants to see a 3D visualization, but sometimes the project size or budget doesn’t support capturing all the measurements you need, replicating the space in CAD, and then working on a design in tools like SketchUp. Since Canvas effectively automates those first two steps, I’ve seen some customers start to just make 3D a default service offering for all of their clients.
- Increasing the use of virtual reality. There are a few different tools to easily take a SketchUp model right into VR. With Canvas, you can now string together a complete pipeline to go from scanning a space, to designing it in SketchUp, to visualizing in VR — all without ever having to touch code.
This all sounds magical. But there must be some limitations, right?
Structure Sensor leverages what’s called a “structured light” sensing system, meaning that it acquires depth by projecting a pattern in infrared light onto the scene. As a result, it can struggle outside, or in rooms with lots of reflective surfaces (like a commercial kitchen). We primarily recommend Canvas for residential interiors for this reason.
Additionally, because we leverage the iPad for processing, you’re currently limited by the RAM of the iPad you’re using - we recommend the iPad Air 2 or higher. For the best results, we recommend scanning one room at a time, and if it’s ever an issue, you can always break the room into two scans and combine the CAD models after.
Lastly, on really large spaces, $29/scan can add up. The average home we process tends to be ~10 scans ($290), and that seems to deliver a strong ROI. However, if you’re trying to scan a 20,000 square foot complex, that can get pricey.
We do have customers that use it for commercial or industrial purposes, as well as some that use it outdoors, but it’s not what the product is optimized for.
What kind of accuracy can I expect with Canvas?
Across case studies we’ve run with professionals out in the field, we see that most measurements are within 1-2% of what’s verified with an existing blueprint or tape measure.
Of course, some projects require higher precision than 1%, which could mean a couple inches across large distances like a long wall. While most customers simply rely on the measurements in our CAD, we do have some that take a few critical measurements (like cabinetry width), and then adjust the final CAD to match what they captured on-site. It still saves them hours vs. doing the whole thing by hand.
You’re obviously very embedded in this space. What else should we know about the state of 3D scanning?
In short: this future is coming, and it’s coming fast.
I’m certainly biased, so I expect to catch some flak for making this claim, but the companies that figure out how to integrate this kind of technology into their process are 100% going to out-compete the ones that don’t. Canvas is just one part of the coming wave of spatial computing that’s going to change the way people make any sort of decision about their home. In five years, I will be shocked if you are able to perform even the most basic home improvement jobs without the client expecting a 3D visualization first.*
Ed. note: Heck yes.
What is next for Canvas and Occipital?
First, we’re investing a lot in Canvas, so the product is going to get better and better over time. I’m sure this audience is no stranger to finding an app on the App Store that demonstrates cool new tech, but it never quite leaves “beta” because there’s no business model to support it. For us, Canvas is a marathon, not a sprint, and we’re in this market for the long haul. Additionally, as Canvas grows its user base and collects more data, we can use that data to improve things like accuracy.
Otherwise, we’re working on a lot of other crazy, moonshot projects in various areas of spatial computing. You can check out our mixed reality headset, Bridge, for an idea of some of the kinds of things we work on.
If you’re interested in helping us build that future, drop me a line at email@example.com. We’re always happy to hear from forward-looking customers attempting to push the boundaries of what’s possible.