patrick.net

WineHorror1 says

There is an app called Cubicasa. This app uses the camera on a mobile device to scan a home in about 5-10 minutes. The file is sent back to Cubicasa to process. The end result is a floor plan with measurements.

Here is a 1-minute video demonstration...https://youtu.be/Iyr4BC1wbm4

The question: How difficult would it be to replicate what Cubicasa is doing?

First thing to test, is if it works. The video doesn't convince me it works well.

From a technical standpoint, this is pretty easy. In the 1990's, they used lidar to produce a completely accurate view of a landscape in civil engineering. This is a little more involved in using video, but it's the same concept and idea. If this was used only by realtors, devices to do entirely accurate mappings of a home have been around for over 20 years. They ought to be cheap now.

Integrating it into a phone? It's using much more basic data input, but it's possible.

Apple has been adding system-wide technology for Augmented Reality. The APIs are public, so you could leverage that to have Apple's code detect objects, distances, angles, etc. in whatever the cameras see. Apple's phones also have a number of cameras designed specifically for distance calculations. Doing the whole thing on an iPhone by leveraging Apple's hardware and AR software would be the easiest route. Still, I would venture to guess the task would be difficult, at best.

If you are trying to write your own code from scratch that runs on a desktop computer with just a video as input .... forgetaboutit!

I'd also try using the Cbicasa system and see how well it does. Don't trust the advertising & marketing people!

Oh... another benefit of doing it all on the phone would be that the software could guide the user instead of just having to take whatever the user happened to record. I could see the software identifying a corner at the floor level and then directing the user to scan directly upwards to the ceiling corner and then around the room in a comprehensive way. The phone also has gravity sensors, gyroscopes, and compass so that orientation and direction pointed don't have to be deduced from the video alone.

WineHorror1 says

What kind of $$ would it take to recreate this app?

I cannot recreate it so I cannot say.

My job is making the ground, and other people make the things on top of it. I like to work on very fundamental levels. I work on the environment, not what is in it. I like to work on creating the dirt and the soil and water. The stuff I work allows people to create everything else on top of it and although I work on the most basic fundamentals, I don't understand what is built on top of it. These are abstractions to me.

Worked briefly for a company that did moving assessments having people upload videos of their home. AI for the item recognition.

I would STRONGLY recommend that you never use such services personally, as the data will certainly be collected by government and larger tech companies.

Huh, I really don't know how they do it.

I would guess it would be somewhat expensive to make.

Your typical programmer / app developer isn't likely capable of making something like this in a short amount of time, much less have it work well. Finding someone capable of doing it might be even tougher considering they would probably see that the market value to the RE industry for such a product is probably high, and I'm assuming there isn't much current competition for such an app, so they might rather build it for themselves instead of you, considering they are the precious resource here, not you.

But the thing with programming is sometimes certain complicated tasks become trivial when someone else does it first, and then makes their code available as a library or service, so others can leverage it. There's a lot of emerging work in the augmented reality space right now, and I think this app is congruent with functionality found in the AR domain. So it may be the case that someone can piggy back on existing code to accomplish the hard part of this, which would drive the price and required skill level wayyyy down.

If I had to guess, I think a way one could do this without adding additional external sensors to a smartphone would be to combine positional data from the gyroscope/accelerometer with the video. i.e. use a computer vision library like opencv to identify features in the video frames, such as walls, corners etc... and then as the user walks a tiny bit, maybe the positional data implies they moved 2 inches, and then you can do some trig to figure out how far in front of you the wall is by comparing the size/pixel width of the wall in the new video frame now that youre 2" closer vs the previous frame etc... I bet I could write code to figure out the dimensions of a wall if it was just a simple square wall in front of me and I was allowed to make assumptions about how the user will walk, but the complexities of dealing with potential objects in the way (couches, picture frames, plants etc...), and irregular shaped walls and rooms and angles etc... would IMO REALLY complicate things. It's one thing to build a prototype that demonstrates a concept, but another to make a quality product that works well in many varied real world scenarios.

I would wild guess people might charge $50-300k for a very basic version. I really doubt you'd find someone who would make it match that apps quality at that price range, as they seem to be a dedicated company with this as their product, and likely have a team of coders working on it full time (each coder cost a company maybe $150-300k yr). Details make a difference, and details cost $$$$$. And IMO, this is the type of endeavor where once you jump in and get your feet wet, only then do you start to realize all the complexities and difficulties of the project. This company could easily have millions into it.

Using some external hardware + software to do the scanning, that then feeds data into a phone, might be a way to piggy back on existing work and drive the price down significantly.

In theory you could have a phone emit chirps and listen to the echoes to find the wall sizes and distances. Big flat spaces will reflect sound well. If you just turned in a circle while doing that, you might get a pretty good map.

Patrick says

In theory you could have a phone emit chirps and listen to the echoes to find the wall sizes and distances.

iPhone 12 Pro and 13 Pro both have LiDAR built into the rear-facing camera group. The 3 optical cameras simultaneously running at different focus distances and the LiDAR are used to generate 3D information. In the camera app's "Portrait" mode, for example, this 3D information is used to focus the main lense and to create fake bokeh. Seems like using dedicated LiDAR would work pretty well! Apple also advertises it as useful in augmented reality, which is the ability to take a real-world video and identify distances and objects — seems precisely what you want.

SunnyvaleCA says

Patrick says

In theory you could have a phone emit chirps and listen to the echoes to find the wall sizes and distances.

iPhone 12 Pro and 13 Pro both have LiDAR built into the rear-facing camera group. The 3 optical cameras simultaneously running at different focus distances and the LiDAR are used to generate 3D information. In the camera app's "Portrait" mode, for example, this 3D information is used to focus the main lense and to create fake bokeh. Seems like using dedicated LiDAR would work pretty well! Apple also advertises it as useful in augmented reality, which is the ability to take a real-world video and identify distances and objects — seems precisely what you want.

Now, at what cost?

Looks like there's an app for that...
RoomScan LiDAR
Most accurate LIDAR floor plan generator I’ve tried a half dozen room scanners with my iPhone 12 Pro, I have to admit that this one is the most accurate. Where other apps have you mark off exact floor corners, which may be blocked off or under counters, this one relies on only marking off the walls. This input generates perfect accurate angles, distances, and proportions. Shelves, kitchen and bathroom counters, cabinets, nothing stands in this thing’s way.

Why would you want to duplicate it? Those apps are off 2 or 3 inches over 21 feet, at best, and over 7 inches with other competitors.
A measurement isn't always frivolous and arbitrary. You either need to be spot on, or over if it's something you're expected to trim down..
I suppose for Real-estate, if all you want is a down and dirty layout plan for rough it ballpark figure, then it would be OK. But then so is scribbling it on a napkin.
And the day turn around, that's pretty lame.
Still nothing beats a real measuring tape, or wheel.
I have made HTML pages, that used Javascript to take room dimension input, and have it make a box Div or Span to that size,(scaled down and in millimeters). Then have them drag and droppable, to where you want them in relation, and click to add doors and windows. It worked well for what we were doing with it.