Interesting data set. I am building a new kind of data analysis tool (https://www.Didgets.com) so I am always looking for good open data sets to download, import into my tool, and see what the data shows and to test out my tool.
I downloaded both CSV files (geometry and simulations) and built a couple relational tables with them in a few minutes. I am confused by a few things. There are 42,207 unique values in the 'apartment_id' column. The most common one is d41d8cd98f00b204e9800998ecf8427e which is referenced 1451 times. At first I thought that it might actually be some kind of 'plan_id' where the same plan was used to build multiple apartments (this id is associated with 13 different 'building_id' values) but drilling down to each one reveals some very different features.
It is certainly possible that the same plan could be used with slight variations (e.g. one has a tub in the bathroom while another had a shower installed), but some of the features were very unique. For example there are 26 different KITCHEN areas associated with the id, but only 21 LIVING_DINING areas.
My tool is great for finding and fixing anomalies in data sets if they exist. This one is a bit confusing about what some elements mean and the site doesn't explain them very well.
If the same plan is being used across multiple buildings, it might be interesting to see how the amount of light entering the building differs based on if the same plan was used to build an apartment on the north side of a building vs the south side.
I think d41d8cd98f00b204e9800998ecf8427e is the md5 sum of "nothing", e.g. md5 reverse will get you a zero-length stream of characters.
(granted this is entirely without looking at the data) but my guess is that they MD5 hashed whatever was in that apartment_id column and if it was empty it spat out d41d8cd98f00b204e9800998ecf8427e
Suggestion: add “is this null”/“is this common?” to your analysis tool. It might take determining that the hashing method is for each dataset “column”, but this kind of trap is everywhere and your users would probably be delighted when they see that’s already identified.
co-author here, thanks a lot for your feedback! We will make sure to clarify on the apartment_id field in the next version. The apartment id is actually the unique id for an apartment in a single site (note: it happens an apartment spans multiple floors) - but it is not necessarily unique over the whole dataset.
We will also include a plan_id field that allows to identify which floors of a building are repeated (the apartment_ids differ already though).
I think the improvements and increased acceptance of prefabricated construction and machine learning can make for an intriguing combination. I am by no means a construction specialist, but if you distill ML to new innovation from historical data sets, architecture certainly has untapped potential.
Just imagine being able to input a geolocation and automatically receiving insight about construction that optimizes for usable space, energy efficiency, or even the prospective homeowner's lifestyle (an AI that recommends different layout options for a family of 5, lifelong bachelor, and non-family roommates on identical quarter-acre plots)
On a slightly more disruptive angle, imagine an AI that could understand a municipality's building code and optimize the space while complying with the literal requirements. Your town has banned finished attics without two methods of egress? Here is an ideal renovation that will provide that necessary balcony while maintaining budget (and here are 4 other buildings in the town that were approved with the same design).
if the genetic algorithm was built such that it used simulated traditional building materials and assembly techniques then sure.
otherwise most genetic algorithms essentially boil down to calculating local minima and maxima of stresses and strains, and optimizes accordingly. The resulting geometry is generally fine for a manufacturing technique such as 3d printing.
Otherwise, despite the material efficiency the labor involved might explode exponentially.
I'm pretty sure I've seen urban planning software that use building codes / parametric parameters to to do this. But architecture / construction has poor history of adopting leading edge tech (constructionphysics.substack.com has great dives into history).
One thing i'm excited for is AI generating ornamentation combined with additive/subtractive manufacturing and we might finally get relatively budget revival of a bunch of more craft based aethetics. Even though we can (relatively) cheaply create detailed geometry now, it's still cost prohibitive to design said details.
As someone who knows Matthias, I can vouch for the engineering effort behind Archilyse's work. I'll admit I was a tiny bit jealous when I first watched their pitch!
They have identified an area where they can clearly add significant value and the analysis their software runs on a dwelling is robustly built and incredibly thorough. I wish them well with their expansion beyond Switzerland!
Probably not the kind of comment that’s usually left on HN but whatever, please tell Matthias (and if he could tell his team) that they went CRAZY with this dataset. As a data nerd I d*mn near started BARKING. Can’t stress enough how HARD they went, I hope their bills are ALWAYS paid, I hope they catch EVERY green light they need to, I hope their pasta dishes are NEVER watery. You get the drift by now. Matthias and gang if you’re reading this thanks for all your hard work. BEYONCÉS of data FORREAL.
matthias here. damn ninjas are cutting onions again. thanks a lot for your kind words, it was an incredible team effort over the last years to create this. we simply hope that the community is going crazy with the data. and there is one more, even bigger thing, we will announce soon. so if this data amazed you, buckle up!
I'm an architect (of buildings) and mostly lurk here but this is very interesting and had to comment - many designers don't like the idea of quantifying design value through measurable data, but it's most definitely the future. The design time and cost savings that can be had from using datasets or software like Archilyse's will continue to grow, while also (hopefully) ensuring a higher baseline design value of buildings. I'm personally just beginning my journey of coding/programming because of exactly this sort of thing (among other reasons).
Since you're an architect, I have a question. Considering the housing shortage everywhere and the land scarcity, wouldn't it be more logical to build higher apartments? My neighbourhood doesn't allow any apartment building taller than 5 floors for example. Why not grant 30 floors? I live in an earthquake zone sure. But there are residences that high allowed in the area, but for middle-class apartments maximum height is limited with 5 floors. I don't get this. Let's build higher, denser apartments and solve the housing shortage. Am I too optimistic?
Your logic is correct, or at least most would say so!
Denser housing could solve a lot of housing issues. The problem is getting denser housing built, especially in not-so-dense neighborhoods. This could be for any number of reasons - the accusatory voice in my head likes to think it's mostly due to NIMBYs ("not in my backyard!" or folks that don't want to live in higher density neighborhoods) because of their warped perception of these spaces: higher crime, higher traffic, unsafe for children, etc. Some of these views may be true, but it's not a given in every high density neighborhood. Safe, walkable, dense neighborhoods exist in many places already.
Another issue can be policy, and specifically zoning, which a lot of NIMBYs fight very hard to control. I'm no expert on zoning, but the general consensus among architects and pro-housing people is that it's holding back a lot of potential homes from being constructed. Check this out for some opinions on zoning:
[https://www.washingtonpost.com/business/energy/can-the-us-ho...]
In my previous city there's a popular type of "middle housing" (not high end single-family, and not small apartments) called a dingbat. They at one point in time were crucial for filling the gap in housing but have now been regulated away. Check this out for info on dingbats: [https://www.youtube.com/watch?v=vlWvcsGlHA4]
For your neighborhood in particular, zoning regulations have likely limited height to 5 floors. Maybe folks in the community lobbied for this regulation to keep density down, or maybe your towns infrastructure can't support a higher density of cars (and this brings up building code and parking requirements that we have in America) - there could be many combinations of reasons for this density limit, but you should look into it! A city's history of zoning policy can be very interesting, as they oftentimes stem from decades old regulations or segregation.
Your optimism is valid - optimism is incredibly important for solving problems like these, especially when so many solutions exist! Implementing those solutions is usually the toughest part.
Fortunately, in California the local politicians have had the power to deny homes taken away from them. The process began yesterday in Santa Monica, where developers automatically obtained permission to build over 4,000 units because Santa Monica's housing element is noncompliant.
San Francisco's housing element should fall out of compliance in early 2023, making it possible to build things without dealing with the local housing cartel.
This is a good question - I haven't studied much aside from densifying existing neighborhoods but when thinking about other options, I suppose suburbs come to mind. Before going any further, I should clarify that I live in the US and look at this through an American lens. Anyways, a primary issue with suburban affordability is that demand is high and space is limited[1]: many folks want to live in the suburbs but maintain their jobs in urban centers, and you can only have so much low density housing within commuting distance to a city. Since urban centers are where the majority of jobs are, it's tough to suggest that people "just move to the country", for example. Remote work could help with reducing density while allowing people to relocate to more remote and/or affordable places. More public transportation may also allow suburbanites to move further away (think high-speed trains and commuter rails).
I've also heard arguments for an urban model that focuses on smaller, more community-centric cities instead of huge urban centers like New York or LA. I don't have any primary sources for this, but I think the idea is to keep density low-ish, and increase the distribution of these urban nodes evenly across a region [https://en.wikipedia.org/wiki/Urban_village]. To me this sounds similar to the mundane suburban towns I grew up around.
Urbanity/density is likely the easiest[2] solution, but I'm sure it's not the only one. There are likely many thoughtful solutions that don't rely on density - I may look around to see if I can find any.
[1] I guess this issue applies across the entire spectrum of housing, which is why there's a housing crisis
[2] Easy is relative - obviously, this is has proven to be a very difficult problem to solve.
LA is close to the logical end of the model you describe; downtown LA is a fairly weak center and there are other nodes as well.
In practice this just means that instead of commuting in long distances in one direction, you do it in every direction. People just tend not to live that close to work due to different desirability characteristics for jobs vs homes. And for multiple income households this is even more difficult, because how often do all the people in a single house work or go to school in the same neighborhood?
i get the impression dense urban environments are ripe for various types of capture not to mention the variety of competing interests that end up diluting the effectiveness of policy to address these issues.
also, i get the impression incentives for housing developers aren't aligned to addressing this problem because in the end it would mean lower margins and a smaller pipeline of future housing development projects effectively putting them out of business.
Keep in mind that, if the barriers of entry are low enough, property development is nothing like a cartel.
Developers will happily undercut each other to steal their competitors’ lunch.
The current model actually promotes cartel behavior; there is so little developable land that it is possible for a few people to hoard the small supply of land.
Oftentimes the 5 floor limit is for timber framed walls. In the United States this is due to fire codes and the ability to evacuate buildings. You'll often see two stories of cement construction for stores and parking with 4 or 5 stories of apartments above, or just 5 stories of apartments alone. Larger buildings are possible, they just don't have the return on investment.
There are some interesting articles on the design choice, but it's a bit sad that we've gelled on this design.
Every era has its “standard” models of construction that people complained about then, and now today people eye watering sums for “historical character.”
The thing that sucks about the current model is that everybody seems to be making retail spaces designed for chain tenants which small businesses don’t need and can’t afford.
This is exactly what city planners call the missing middle in the US. There's skyscrapers, and two blocks over you have single family suburbs.
If it wouldn't be for political opposition most cities, undoubtedly, would upzone (permit higher density) areas near downtown and other hubs, but even in "progressive" cities progress is slow because of opposition of the people who currently live there (commonly called NIMBY's)
Archilyse provides a SaaS tool to convert floor plans (raster images) into 3d models of buildings (IFC or GeoJSON) with embedded contextual information. the 3d model accuracy gets verified via additional data sources (governmental building hull data, client database entries) and all buildings are geolocated. subsequently, in a 25cm grid multiple simulations (3d view shed analysis, daylight, traffic noise, centrality, discrete metrics) are computed and aggregated into feature vectors. these are used for training AVMs (reducing prediction errors in half), architectural analysis (judging architecture competitions), construction cost estimation, life cycle cost estimation, energy load peak prediction, ... different research groups use this dataset to derive design patterns and to come up with augmented ai workflows for architectural design (like copilot) or as benchmarks for novel fitness/cost functions in MCO strategies.
The way we shape our maze determines our behavior which one could use to design personality. For example, if the kitchen is on the road side you wont see your neighbors as often. Or 2 bathrooms give your grumpy morning mind some alone time.
In a way we are different people in different rooms. The transitions could be interesting to explore. For example toilet > kitchen is not done or even illegal. A toilet in the garden on the other hand seems fun.
Maybe hybrid rooms build an interesting character. Say a bath in the middle of the living room next to the fire place. A kitchen library also seems fascinating.
One question has been bothering me: why are rooms square?
Some say that it's more economical to build this way, and that it causes less problems with aligning furniture to the wall.
But that doesn't explain billionaire houses. They certainly love spending money on them, yet despite all kinds of extravaganza, the rooms are also mostly square.
So i think it's something deeper, I think it's just too suspicious that across most modern cultures rooms are square. My theory is that it's related to the fact that we have 4 sides, so it's kind of symmetrical that we prefer to live in 4-sided things too.
High-end houses aren’t square in the same way that a normal house is square: often they’ll be open plan, where rooms join together through shared spaces that are all sorts of shapes. A better description of the consistency amongst property is “straight” walls.
If you spend much time looking at high-end real estate, you’ll encounter much more than just the standard square rooms you would see in the average house, but ultimately, they’re still square(ish) because straight walls are convenient and practical.
I think the best description would be “right angles”.
Ultimately, not only is furniture designed for right angles, but so are all the construction materials and equipment. Doing anything else is custom and more expensive but also not usable because of the furniture limitations.
The best you can get without going into custom stuff il are those bay windows that pop out.
pg wrote about this a couple years back. big-boxy architecture of american suburbia vs all the diverse architectural styles in europe/asia etc.
in many parts of costarica, panama & especially in upscale indian houses, the house looks more like an art project. i’ve been to houses with what would be called gaudy colors - blood red, bright yellow, parrot green walls. non-square rooms, oval spaces with arches and round pillars inside the house. in those places, people use their art skills not just on canvas but also in living spaces. it might not look conventionally pretty, but everybody gets a say - like in the kitchen, my mom had holes of different shapes in the wall, so bottles of different sizes fit into specific holes. no actual shelves! i wanted a table so i got a table built of concrete and cement! just put a tablecloth on top. no need to buy a table, table was an actual part of the house itself. i’ve been to bathrooms with a raised mound of cement to hold pots of water. you carry water from well and you can place the pot with water on the raised mound without bending down all the way to the floor. very thoughtful ideas.
Furniture is pretty difficult to make for for rooms with odd wall configurations. You have to custom design every piece and wind up losing a bunch of space in a dresser or closet in most cases.
seahorses have 4 sided spines. parent poster is of the pipefish family, and their question is not about human housing, but frustratingly rectangular aquariums.
This is a nitpick that's not related to the article itself, but I found that sorting by most viewed, the asc/desc dropdown has an effect from what I expected.
The intersection of architecture and 3d design with ML is an intersect I haven't seen much work in, but it would be fascinating to see what comes from this
I downloaded both CSV files (geometry and simulations) and built a couple relational tables with them in a few minutes. I am confused by a few things. There are 42,207 unique values in the 'apartment_id' column. The most common one is d41d8cd98f00b204e9800998ecf8427e which is referenced 1451 times. At first I thought that it might actually be some kind of 'plan_id' where the same plan was used to build multiple apartments (this id is associated with 13 different 'building_id' values) but drilling down to each one reveals some very different features.
It is certainly possible that the same plan could be used with slight variations (e.g. one has a tub in the bathroom while another had a shower installed), but some of the features were very unique. For example there are 26 different KITCHEN areas associated with the id, but only 21 LIVING_DINING areas.
My tool is great for finding and fixing anomalies in data sets if they exist. This one is a bit confusing about what some elements mean and the site doesn't explain them very well.
If the same plan is being used across multiple buildings, it might be interesting to see how the amount of light entering the building differs based on if the same plan was used to build an apartment on the north side of a building vs the south side.