RoomScan: Get a Floor Plan in Minutes Just By Walking Around the Room | Hacker News Reader

RoomScan: Get a Floor Plan in Minutes Just By Walking Around the Room

architizer.com

230 pointstudborg75 comments

Here's an alternative with a more interactive work flow: http://magic-plan.com/ .

That seems to work a little better for me in terms of accuracy. This approach is different in that this is using a fixed eye position and overlay measure on the camera image and then working out the arctangent etc. Clever.

Considering I'm only using it for getting a general layout of rooms for my wife and I to place furniture, this is awesome.

Tried it. More bells but difficult to make accurate. Hard to use if you can't see the floor corners of a room or if dealing with short narrow spaces.

As did I, quite some time ago. I found it did a poor job with rooms that had actual stuff in them, -particularly the corners. Seemed cool but I discarded it as an option even without an alternative. RoomScan has been making its way around the blogs and I tried it out the other day. It did a great job of capturing my non-rectangular kitchen (after I failed at my first attempt by ignoring the instructions to touch each wall, then the 1st and 2nd wall once more, then ending). Roomscan > for spaces with stuff and/or narrow spaces, or Magic-Plan > for empty spaces more wide open was basically my experience.

Easy enough to try both though.

Quite disappointed - was hoping this was done through computer vision. Aren't there are apps that do this? By the time I've 'scanned' a room the way it's shown in the article, and if that method doesn't even give me heights, doors, windows and outlets, and is only accurate to within half a foot - by that time I've measured it with my laser tape measure 3 times over.

I don't think we're quite at the point where a mobile CPU would be able to pull this off, however filming and uploading a well shot video to a server for further processing should be a workable way to make a full model. I think a big challenge there is a several-minute feedback loop to learn if it worked and if you need to try it again.

I built an Android app, Forge, that does just that. No feedback loop needed - the reconstruction is streamed back to you as you go:

http://aboundlabs.com

Incredible demo! You in the bay area?

Thanks! I was - moving to NYC Monday.

Also very impressive -- just posted

https://news.ycombinator.com/item?id=7474680

Is monocular SLAM good enough nowadays that this can be done without depth cameras, just from a single camera?

I was talking about monocular SLAM (although since it's not really simultaneous... ALAM?)

Parent has a cool product too, as does http://structure.io/ - but that's not what I meant.

Probably for some version of done - download VisualSfm and give it a shot yourself if you want: http://ccwu.me/vsfm/ or for a purely mobile approach try http://seene.co/

And your laser tape translates those measurements into a floorplan that's digitally sharable?

I wonder when the trend of "X, but shareable" ideas will stop.

His example didn't even yield X.

A digital (and therefore shareable) floorplan is plainly more useful in a variety of ways than one that exists only on paper

A smartphone camera very quickly turns a non-digital floor plan into a digital one.

It could send a photograph of a floorplan, but it wouldn't be editable unless there's some interpolation of the image you photograph. When your original floorplan begins in digital form, it's more-flexible for editing.

no,. the sharks do that bit.

My HOWTO:

1. Obtain laser tape, pencil, and paper.

2. Sketch out floor plan with pencil on paper in rough scale.

3. Take laser tape and take needed measurements. Annotate measurements on your rough scale paper drawing.

4. Go back to desk and open OmniGraffle (or Visio or whatever).

5. Draw up a scale floorplan, exportable into any needed, industry standard formats.

6. There is no step 6.

Some variation of this HOWTO has worked for me professionally for 2 decades. Sometimes, the old way is better than a novel, high tech solution.

Edit: And yes, I did try the Roomscan app. It was tedious and inaccurate.

I bet someone once said the same thing about the laser tape replacing traditional tape measures.

Until they tried to hold a tape measure straight to measure a 30-something foot room...

This argument depends on specifics, not the overall form. Just because another argument with a similar form was made and it was wrong doesn't mean anything about this one.

No, absolutely nobody ever said that. The only con of a laser measure is the price (compared to a foldable analog one), but apart from that (and for the purpose of measuring rooms and buildings), a laser measure is superior in every way.

Has that been true from the beginning? It's very rare for the first iteration of a new technology to immediately be superior in nearly every way.

"No, absolutely nobody ever said that."

http://www.woodweb.com/knowledge_base/Taking_Measurements_on...

You can't see the laser dot in sunlight at a distance > 10m and for some measurements keeping the point steady is important but hard to do without a tripod. In both cases the laser can be inaccurate as you can't be confident in what you have measured. Obviously laser theodolites can solve these problems but are about 1000 times more expensive than a reel tape.

How to start a startup: Find a process that requires tedious work with lots of different tools, and make that process simpler and easier. Start with something that works well enough to be of value to some people, and improve until the old process is laughably complex.

This app seems valuable to non-professionals who may want a digital floorplan of a room. Since you are a professional with 20 years of experience, you have an established workflow, and know how to deal with the variations.

Experts will often be able to out-perform automated solutions, which is fine. Those automated solutions are for non-experts.

I don't think you're giving his argument enough credit. Since he's a pro, he probably knows what he's talking about.

I've tried Magic-Plan, and yeah, it was frustrating and too inaccurate to be useful.

Start at the use case: Why would you need this? As a home-owner who's done it, I can't think of any except remodeling. Wether you're replacing the carpet, getting rid of a half-wall, putting in a new beam, measuring out a bathroom for a remodel, you need good, accurate measurements for this stuff.

If just eyeballing it was good enough you wouldn't need the tool in the first place. Guess. Look at your 1' floor tiles, count 'em, and you've got a probably better read on everything.

What would've really helped me out when doing this was a great, simple to use tool to help me build floor plans. Most of what I found on the App Store for the iPad was pretty shoddy, or just difficult to use effectively. Like I want snap-to-grid. But once I put a wall in the exact location I want the option to lock position. Or lock length.

So I dunno. I feel like he/she has a point. When approaching this problem, and needing to make a compromise, you don't compromise on accuracy or you've made your tool irrelevant. If they'd just focused on a really great floor planning tool that would have been the better compromise.

That said I haven't tried it yet, so maybe that's there. I'm gonna give it a shot.

floor plans when finding renters, determining furniture layout, how much paint to buy. to name 3

> This app seems valuable to non-professionals who may want a digital floorplan of a room.

No data is better than bad data.

Here's a high-tech method that usable in a professional setting: http://www.theage.com.au/technology/sci-tech/melbournes-shri...

Would you be interested in a service that takes your paper drawing and does step 4 and 5, in Autocad and gives you the results within 24 hours?

Cost would be at $10 - $15 depending on house size.

For only $15 ? I know a few people doing CAAD technical drawing, all day long. Often just recreating the sketches the structural engineer draw on a piece of paper.

Are you offering to do that manually (prolly outsourcing) ? You know, those paper drawings aren't 4 lines with 2 measurements :-)

Or are you looking into automating that ? That would be a very interesting side-project ... but that would be really hard to do character recog. on hand-written numbers that are all over the place :-/

Sounds like an opportunity for an app ..

This is probably what you're looking for:

http://structure.io/

And while not germane to your particular use, this is still worth noting:

"Structure allows you to capture dense geometry. This enables you to simulate real world physics."

Would Leap Motion or Kinect be able to do something like this? They basically detect physical objects in space so shouldn't they be able to 3D map a space? I know the current Leap device has a limited view area but I understand that's just a design decision for that particular device.

I was a residential contractor until a few years ago. This would be great for estimating. Nice way to replace a pad of paper with less than stellar drawings on it. When estimating I'd still take pictures of the rooms to help me recall what it looked like.

For most projects a measurement to within half a foot is fine for estimating. If the total width of a room is close to 12 or 15' (standard carpet roll widths), get an exact number to know if your carpet sub needs to seam two pieces of carpet together.

Door heights are standard on most houses. Electrical heights are usually standard. Windows are usually dropped the height of a header.

For estimating many projects (painting, flooring, non-custom trim jobs), this app would be more than adequate.

When you get the job and go back to do the work, I'd suggest a Bosch LASER tape measure. Accurate to within 1/16" or better. It's great for trim work.

Yes, I have a Bosch. From the looks of it, this app isn't faster, nor does it add any other significant advantage. You still need to measure the location of windows and doors, width etc. And you need to manually add that to the plan. Or, what I usually do (although I'm a developer, not a contractor), is write down areas directly. I'm still not convinced, and if they're targeting the professional market, I am part of their target group. YMMV etc., obviously, and if they manage to build a business out if it, more power to them.

What I imagine to be a real productivity boost would be a combination of both: A laser tape measure paired via blutooth to take exact measurements and an app using the accelerometer/camera for the general outline. In a best case scenario the app could even locate the laser tape measure to detect which measurement is taken.

Very nearly. Google's project Tango. After you are finished oogling at real time SLAM using vision running on a phone, I invite you to check out Google's Project Ara. Both were spawned from Motorola ATAP - post google aquisition - and retained by Google after the sale to Lenovo.

This seems to be in the neighborhood of Johnny Lee's thinking in Project Tango: https://www.google.com/atap/projecttango/ ... a lot of the information collected by the increasingly precise sensors is going to allow for next level of... something... I think Project Tango is an example of not knowing what the future is in sensing and looking for network effects to fill in the gap.

Tried this app out a few days ago. The measurements in it are surprisingly good. I was finding it about ~10% out at most.

All in all, it is a very good (FREE!) app that can be used to get a decent idea of what might be needed in home renovation.

I had the opposite experience. It seemed really finicky about how smoothlyI had to move my tablet from one wall to the next .

Yeah, I couldn't really get it to work either. The preview seemed to be more or less accurate, but it wasn't able to generate the layout. I was trying to double tap to modify the dimensions manually, but it looks like the free version doesn't allow for it. I guess it's a hit or miss.

I hoped that either they got their hands on Project Tango device or worked with this team in Zurich - http://www.i-programmer.info/news/192-photography-a-imaging/...

Edit:

Btw, anyone who applied for Project Tango got back from them in any way? I didn't.

I applied with an idea very similar to the OP, except I'm interested in capturing GIS coordinates and video to automatically capture engineering drawings for central offices, head ends and data centers that include racked equipment.

I haven't seen even an e-mail confirming they received my application, so I don't know whether I was rejected or never entered.

Structured light cameras like Project Tango or Occipital's Structure sensor that project a textured pattern would work well here. Stereo algorithms like the one you linked that use the regular phone camera are likely to struggle here, because flat, blank walls are a worst-case scenario for them (yes, they're using the accelerometer to assist, but the reconstruction still needs texture).

We heard back and got a device last week.

If so, they're being rude toward the rest of us.

They certainly had a ton of responses, and possibly they plan on working their way down a list or something.

Could anyone possibly shed some light on what is used for measurement? gyro?

Yes probably, but I would be surprised if it isn't also integrating over the accelerometers' output. See http://en.wikipedia.org/wiki/Inertial_navigation_system.

I was actually beginning work on a similar project. Essentially, it's using a combination of the accelerometer and the gyro to get the phone's position in 3D space. That is, when the accelerometer capture lateral velocity, it knows it's moving in a certain direction at a certain speed (until, of course, it receives an opposite velocity, at which point it stops).

Did you ever manage to get it to be accurate, though? I tried something similar but the accelerometer and gyro were nowhere near accurate enough.

Supposedly, there's a rule of thumb that says that dead reckoning by integrating readings with a MEMS sensor is about as good as walking around with your eyes closed. I didn't even get mine working that well when I tried, but I'm sure it's possible if you get the pre-integration noise filtering right.

Lately I've seen a lot of app notes and open-source code written by/for amateur drone guidance developers, which I suspect would have helped me a lot.

It's unlikely to be standard inertial navigation, mems sensors drift with dead reckoning integration pretty quickly, even applying more sophisticated kalman filtering may not improve things indoors without GPS (or sufficient travel for GPS even if you could receive it consistently)... But, if you were to add "room-like" assumptions it might improve things pretty well, eg. when integrating a rotation, fit it to a 90 degree corner, similar with traversal, fit it to travel along a wall. With those basic assumptions, I'm guessing you can correct a lot of the raw integration drift. If this is true, my guess is it would start break down in unconventional rooms - octagonal rooms, non parallel walls, etc...

Bingo. Here's an example from a room with a centered bay window: http://images.younghouselove.com.s3.amazonaws.com/2014/02/Ap... So you have to go back with a measuring tape and amend all the measurements anyway, because it's so inaccurate.

Full article reviewing MagicPlan and RoomScan: http://www.younghouselove.com/2014/03/apps-and-zerts/

I wondered about doing something like this a while back - and it was the very quick cumulative error from double integration of acceleration data that stopped me from even trying. Clever idea getting the user to place the phone on the walls - as you say that would correct a lot of the errors. But also, as you say, unconventional room layouts get you back to square one so to speak.

I haven't played around with the latest generation of sensors, but if the time/distance error growth is small enough, maybe you can get model corrections down to the nearest 45 deg. If you can hit that threshold, you might end up fairly functional with possibly the next largest segment of non-rectangular rooms.

Not even close. I was approaching the filtering mechanisms before I started a new job and started focusing my efforts on that.

Try this http://www.cs.huji.ac.il/~gaddi/3D%20Motion%20tracking%20wit...

How does it measure the length of the wall by holding it in just one spot in the middle?

Google's Project Tango (https://www.google.com/atap/projecttango/) and Structure.io seem to be the most promising in this area.

I have difficulty seeing any professionals adopting RoomScan as it still seems more difficult than just using a laser measure.

This would be nice if I can feed the output to my neato( or roomba).

I got a lot more excited when I read about this a while ago (2012) [1]. It also states to be slightly more accurate. It also seems less cumbersome.

[1] http://www.extremetech.com/extreme/126843-think-gps-is-cool-...

I haven't tried anything, but it seems there are already lots of solutions and this is just another approach.

AppStore link: https://itunes.apple.com/gb/app/roomscan-app-that-draws-floo...

Couldn't find it on my brief scan of the article.

I've tried this a while ago. Did not work very well at all on an iPhone 4S, it would either tell me to re-try or come up with a plan that was completely off.

"Your phone wasn't able to record all of the movement. Please try again"

can't get past 2 walls. iphone 4s

HomeDepot should let people use this instead of charging 30 for a measure and loosing it 3 months later

I usually Look at it for a few seconds and then think about what I want to do in there in my Brain.

Does anyone know how it actually works?

Reply on news.ycombinator.com