Hacker News Reader

Comment by guidoism | Hacker News Reader

I tried hard to make Sweet Home work for me. It solves the quick and dirty use case but I really wanted something where I could measure out every dimension in the house with a laser measurer, describe the shape and the relationships between the objects and let the computer do the rest. In a program like Sweet Home, and every other floor plan program out there if you make one change (like updating the measurement of the wall thickness you have to manually and carefully move everything else around.

I ended up throwing together something quick and dirty with Org Mode tables and Metapost: https://github.com/guidoism/wildwood/blob/main/house.org

It works pretty well and the output is pretty.

View on HN

That sounds fantastic, and it mirrors my (brief & shallow) experience with such tools. It quickly becomes too much of a drawing project, and too little of a data-capture/tabulating one.

But please please please show the resulting picture somewhere, it was incredibly frustrating to read through your page's wonderful build-up towards the goal, and then not get to see the rendered result? Poor me.

https://github.com/guidoism/wildwood/releases/download/20230...

What I ended up doing when I was in a similar situation is manually edit the XML(sh3d files are XML) with proper measurements and coordinates after first doing a rough sketch on the GUI.

I haven't used that tool for a while, but I think that it includes a table with all the items (walls, etc.) where you can enter the measurements.

Ultimately I think of sweethome3d as a sweet spot between paint and a full-blown CAD program. Nowhere as powerful as the latter but so much better than the former. Also my daughter had a moment when she preferred sweethome3d over Minecraft et al. as a building game.

+1 on the sweetspot. I never mastered a proper CAD tool. Sweet home 3D was amazing at helping plan the furniture build and layout of an empty 40m2 living space. I have been living in that space and benefiting from the simplicity SH3D for ten years now with no need for a change.

This has been a source of frustration for me too, as I've occasionally tried out these kinds of packages. Particularly having used model-oriented dimensionless CAD tools like Solidworks, it feels like stepping back in time four decades to suddenly have to manually dimension and align everything upfront.

I modelled my whole house including very piece of furniture. It is very “finicky” as you describe. It’s good enough for my amateur use with manual clicking. (There’s probably a programmatic option that I don’t know). However if I was a pro contractor it would probably not be the best.

I did plan custom designs that my contractor used. It’s also great when I want to get a new desk or sofa or whatever so that I can visualise everything first. Possible to make pretty realistic 3d walk-through videos that would be fine on any design tv show.

I shared the same frustration, but I finally gave up, and guess what, I was able to fulfill my projects and have a good insight about what my layout will be even with ~5cm difference. I learned to relax about absolute measures and think more about feeling and subjective dimensions.

Described in programming terms: You try to type the weird dynamic object you have to deal with in Typescript but after spending too much time, you learn to relax about types and just write it in JS and it feels like it's working most of the time.

Don't forget the power of `as const` and type guards for tricky dynamic types! Especially for function params that accept multiple complex object types.

Parametric modelling - OpenSCAD (which you mention) is commonly used for that style of handwritten arithmetic and has a few python-based alternatives like CadQuery. In GUIs, FreeCAD has an architecture module that aims to be fully featured (never used it), but for just parametric drawings, SolveSpace is lightweight and probably more pleasant to use.

Have you tried LibreCAD? It seems specifically designed for 2D, but I don't know how "parametric" it is.

OpenSCAD is completely the wrong tool for the job. You want a graphical parametric modeller. The only vaguely decent FOSS one is SolveSpace. It's a bit lacking for actual CAD (no fillets!) but it works fine for 2D layout.

I have actually used it for that (actually I did full 3D modelling of part of my house) and it worked ok, however I would say that measuring things with a laser and copying them into CAD is not a very accurate way to do things.

You will end up with large accumulated errors, and most houses are not as square as you think they are.

I haven't ever used one but I suspect those phone based AR measurement apps have a good chance of being more accurate.

> and most houses are not as square as you think they are

I'd say most houses don't even have straight walls, so you should be very happy if what you got is vaguely a trapezoid.

I wonder if there is a good answer for quick mapping using an iPhone’s LiDAR Scanner?

This is definitely the eventual answer. I can't imagine in ten years we will be hand measuring walls.

Oh, yes we will.

Unless you want to move all the furniture out of place (even in-built one), all the heavy stuff blocking edges and corners from view etc. Good luck with that bathroom sink or kitchen counter-top.

And angles and inequalities (eg. top edge vs bottom edge), that gets funny fast.

isn't that all of the more reason to use a high-resolution 3d LIDAR point cloud - to get the exactly geometry? I hear your point about moving furniture, but you kind of need to do that anyway to get a measuring tape into corners.

Couldn't ML be utilised to recognise furniture and remove it from the model?

Laser measuring tools avoid the need for a physical tape these days.

Mostly!

But then you need to measure "free standing" wall (when measuring the opening is not an option), and without someone to hold something flat at the other end for the laser to bounce off, measuring tape still wins.

Which is not to say that laser is not extremely useful: I generally use both.

Measuring tape is pretty thin and can get behind or just under a bunch of things: or a laser can when tape can't.

Basically, my point is that we'll need all three, and workflow won't be trivial to combine them.

There are companies today like Matterport: https://matterport.com/ which leverage LiDAR to do high detail measurements and mapping for various applications.

My company uses them to measure prior to use creating build plans for remodels and I used them at my previous company to do high detail measurements in the context of real estate to make 3D tours of homes.

If the iPhone can eventually get to the point where Matterport is today with their tech then that's good enough for a huge number of use cases.

Pretty much everyone in the construction and remodeling industry right now is prepping for that as the future. We all know it's 3 to 5 years away and want to be the first one to get it right.

Everyone is prepping for those self driving cars for the last 10 years, yet we are still swinging those steering wheels ourselves :)

There are certainly cases where LiDAR approach can work wonders. But as long as you need to do a couple of corrections manually, you'll need to have both set of tools at your disposal.

It will definitely augment current measuring approach (or rather, existing tools will be used to augment results coming from 3D scanning), but as soon as you have to pull out a laser or tape for one edge, your workflow is significantly more complex.

Not to mention that for cases where one might use SH3D, it'd be hard to tell an automated tool to ignore that 2"x2" "tooth" in one corner for just wanting to look at different furniture arrangement.

TBF that would be a good application of AI to infer what is clutter and what is not from the visual light image of the same space, the wall behind an object can simply be inferred from the sections of that wall that aren't occluded by the clutter. It might not catch some obscure cases where clutter obscures some anomalous part of the wall, but it'll be good enough in the other 99% of cases. Multiple viewpoints would probably deal with a large number of oddities too - much how photometry can get more accurate with more viewpoints already.

I found this youtube channel[0] has some pretty good walkthroughs (quite literally) of using various iOS apps to scan architectural spaces. The short answer is that the LIDAR and iOS APIs are remarkably powerful, but not 100% accurate. There are techniques to improve accuracy (e.g. using a gimbal), but ultimately you'll need to do tape or laser measurements and modify the models that these tools can build, or just model it yourself with the scan as a reference.

MagicPlan[1] and PolyCam[2] seem to be the most focused on building a schematic level building model which could be imported into other tools if needed. They both now take advantage of the Roomplan API[3] which Apple introduced in iOS and iPadOS 16[4]. MagicPlan has been out for ages[4] and originally just worked off the camera and the accelerometer to help build a floor plan. Polycam also supports photogrammetry[5] where you just take a bunch of photos and it builds a 3D model by interpreting what shape the object could be (I don't know if this is also used in architecture scale things, but it could be interesting for ID projects). Both MagicPlan and PolyCam allow you to tweak dimensions of rooms, doors, windows, furniture, etc. in a somewhat parametric way. This is where you likely want a laser measuring device to quickly update the dimensions. These can be used through Bluetooth to enter the measurements directly into the floorplans in MagicPlan[6]. I didn't try this, but if I was doing this all the time, it seems like it would be essential.

Matterport is starting to get into mobile[7] (phone, tablet) capture, but they've built their business up on their branded hardware and cloud platform. They provide floorplans as a service[8] and everything adds up, but from what I see in the real estate market, they are ubiquitous.

And if you want to spend a bunch more for very pro level app for documenting things like crime scenes, shipbuilding, infrastructure, etc. there's Dot3D.[9]

[0] https://www.youtube.com/@LiDAR3D

[1] https://www.magicplan.app

[2] https://poly.cam

[3] https://developer.apple.com/augmented-reality/roomplan/

[4] https://9to5mac.com/2022/06/15/ios-16-roomplan-api-3d-floor-...

[5] https://www.magicplan.app/about

[6] https://help.magicplan.app/laser-distance-meters#laser-tutor...

[7] https://matterport.com/3d-camera-app

[8] https://buy.matterport.com/

[9] https://www.youtube.com/watch?v=ouZxCDKTizs

There are several products out there that can create 2d, 3d renderings of a living space using the iphone lidar scanner. Not sure it they keep dimensions.

I have been playing with RTAB-Map and a Kinect, and exporting to various formats and it works reasonably well, but is still far from user-friendly. They also have a iOS version if you want to play around with it.

Ooh, I just found my old Kinect yesterday. I'll have to play around with this.

Download RTABMap and the OpenNI drivers and export as ply and use CloudCompare and MeshLab to play with the pointclouds. I am not a fan of the mesh function in RTAB-Map and there are some buggy options which insta-crash it on export. Meshlab is also super-buggy and has a terrible GUI. Have fun!

EDIT - PLY not PCL

I work on an app called Polycam, and we have an automated Room / floorplan capture mode. We actually announced some updates to it earlier this week:

https://twitter.com/Polycam3D/status/1623730477637959680

thanks, this is super cool. Does it have decent geometry export options?

Yep! For meshes we've got GLTF, OBJ, FBX, STL, and DAE. And then you can also export the 2D floorplan as SVG and DXF.

Love Polycam! My wife just got the newest iPhone this fall before our remodel started and I was stoked to capture our house before the remodel and cannot wait to get the after all scanned in for comparison. https://poly.cam/capture/11B48FCD-E6A5-4ED7-9C9D-BF5DFE60579...

I love the app! The interior designer and property agent who saw me using it were pretty blown away.

One thing I struggle with though is making sure that the capture is robust enough before I move on. I usually only really know that something is off when the capture is done already. Can I “touch up” an existing capture by doing another local scan of a problematic area?

You can - there's an 'Extend' button for LiDAR captures, but it can be a little touch-and-go sometimes. Definitely an area for us to work on more!

I have an app for creating large scale, drift free scans using iPhone: https://imagespace.app

You (like me) are looking for a parametric modeler. SolveSpace is pretty good for that

>You (like me) are looking for a parametric modeler. SolveSpace is pretty good for that

OpenSCAD is my go-to since you just write code.

I never figured out how to relate primitives to each other in OpenSCAD so my models were an endless soup of absolute position math. I used SolveSpace for a while and it was brilliant because what took hours in OpenSCAD took just a fraction of the time. And then I picked up the Maker license for Solidworks ($100/yr) and haven't looked back.

I rendered my whole house in Solidworks to plan for a remodel and it was extremely well suited for the job.

I'm going to try SolveSpace, but I'll look into that Maker license. Didn't know of that, and I can absolutely afford a grande-latte/month for a tool like that.

The nice thing about FOSS though is that it's worry free. I can keep OpenSCAD and Cura, and my projects will build years later without me having to think about it.

The stuff I built with 3DS Max educational license, though, is not accessible. So I am hesitant to jump ship for Solidworks (even though it's the industry standard).

I come to HN for comments like this.

It’s been awhile since I’ve used this, but IIRC you need this plugin to be able to easily update dimensions without manually moving everything around: https://www.sweethome3d.com/support/forum/viewthread_thread,...

Ugh, thanks. I was hoping it'd be more parametric based, like Fusion 360, but alas seems not.

This was one of the unexpected extra benefits of having a robot vacuum with LIDAR, free accurate floorplan after one sweep.

Obviously you just get an image, but for what I needed (map in home assistant with sensors overlaid) it was enough.

The vacuum navigates and route plans based on the geometry, though, so chances are the actual map is on the device. Robot vacuums are competitive and plentiful enough to where there has to be one with a way to download the map from some endpoint on the device, or at least a way to pop it open and pull out some SD card. Of course, it’d probably be in a proprietary format, so transforming it into a usable mesh might be tougher than getting the data.

Those scanners are IIUC typically line scanners, though.

There was a startup in SF that tried to do that using iPhone video feeds as one walked around the room, but they went belly up recently and whatever tech they created was siphoned to another platform startup.

So few people use Metapost, it's undervalued! Feature request: add the output to the github repo so we can see the result?

> org-mode

Of course you did....

Gods above I love org-mode.

Reply on news.ycombinator.com