Show HN: Using YOLO to Detect Office Chairs in 40M Hotel Photos | Hacker News Reader

Show HN: Using YOLO to Detect Office Chairs in 40M Hotel Photos

222 pointsnomad86124 comments

I used the YOLO object detection library from Ultralytics to scan over 40 million hotel photos and identify images with office chairs. This helped me create a map showing hotels suitable for remote work.

Map: https://www.tripoffice.com/maps

Yolo: https://www.ultralytics.com/yolo

The whole process was done on a home Mac without the use of any LLMs. It's based on traditional object detection technology.

View on HN

This is like an automated version of 'Commercially Available Chairs in Star Trek' but for hotels. https://www.ex-astris-scientia.org/database/chairs-trek.htm

This will come in handy to be able to perform the Riker manoeuvre:

https://youtube.com/watch?v=lVIGhYMwRgs

IMHO There is a huge gap for affordable hotels for normadic workers and business people.

Hotel rooms suck when you need to use them for work. Typically there are massive beds and I travel alone if for work. There is no proper chair, no writing table at all or one too small, and the sockets tend to be in the wrong corner of the room.

If I was an entrepreneur outside of software looking for a gap, I might have invented a hotel chain for work stays. But I'm not, so here is the idea for you to get rich with it (so I can stay there one day).

I like the OP's idea of using ML models to gather intelligence from hotel photos. For years I took a photo of nearly every hotel room with my laptop on the desk so that I could go back and re-book the rooms that were suitable if there was a conference in the same city again in the future.

Once I specifically booked an airbnb because they promoted their 17" LCD with hdmi, usb-c and dvi inputs. I'm like, why don't more offer this?

I think the answer is the first word in your sentence. How many times have you stayed at a hotel? Now how many times have you stayed at a hotel because they promoted a good work space? Both cost about the same to run, and one has a much greater need.

Plus people that do like to spend much of the year travelling around doing deep work are (i) relatively likely to book short stay apartments instead of hotels and (ii) relatively unlikely to be particularly fussy about working environments since they're actively choosing travel over convenience and optimal working environments

If I think I just can't work without multiple monitors and a high-end office chair (and maybe printer), I probably won't travel or I'll get a co-working space of some sort. When I traveled a lot for work, it was some combination of the event/trip was my working and/or I just worked on my laptop wherever.

I'm semi-retired now but I'm temporarily staying in a Marriott property (Springhill Suites) that does have a usable desk and office chair which is just fine for writing at for me--though people with very specific requirements probably wouldn't like it.

Plugging in random USBs is brave!

Depends on where you're looking.

There are usually plenty of "business" oriented hotels near airports, business parks, central business districts, convention centers, etc . (And that definitely reflects on the trip office map). Touristy areas have more tourist/traveller oriented amenities.

The one with a room they call business center which has a printer and a fax machine?

They know what they’re doing. Most hotels ‘are’ built for work, but they’re built for what most workers use them for: work travel. Though probably common for developers, I’ll bet it’s pretty unusual for the other like 96% of people to want to do deep work in a hotel room. Especially if you’re on the road for work, you’re probably the sort of worker that needs to physically be somewhere remote to accomplish a goal, which makes it even less likely you’re going to be doing deep work in your hotel room. Most have “business centers” where you can bang out a zoom call, rally the troops before a meeting you flew in for, print something out for a presentation, or whatever. I’m sure most professional use cases are far better served by those accommodations than amping up their rooms for the handful of people that need to be in a room by themselves sitting in one spot for long enough for the chair to be a big factor while they travel.

Hanlon's Razor

I think there's a simpler explanation: most people don't do work when they go to hotels, they do the work like you are discussing. Doesn't mean they are intentionally being hostile to remote workers.

One thing I've realized about the world is that a lot of people do things just because others are. "Momentum is a bitch." I will bet you very few people are thinking this way, at least very few that make decisions. And the ones that do probably think it is not worth the money. There's a ton of things where markets don't exist simply because the environment doesn't exist, so the people that can make the environment don't because there is no market. It's the whole "build it and they will come" thing. People are very risk adverse. People are hard to move. Would hotels benefit from this? Probably. I mean even not just considering nomads, most people work from their computers[0].

But it very easily could be one of those things where there's push because there's no market and there's no market because there's no push.

[0] The way people have been talking about working at CES has sounded silly. There was a LTT video where they mentioned how WiFi used to be better in some locations so those rooms were more desirable and the hotel's solution was to make it standard for everyone. They seemed to be suggesting that they brought down the quality rather than balanced.

I’m not sure how what I said could be interpreted as suggesting they are hostile to remote workers.

And I think there’s a very good chance hotels would not benefit from this. Maybe in a tech center, but that’s a tiny fraction of hotels. Good office chairs are designed to be very adjustable, but they do tend to break when people twist one thing too far the wrong way because they don’t know how it works, and it would probably take staff 10 minutes just to figure out it was broken rather than just misadjusted. They’re also expensive as hell, and charging a guest $3000 because their luggage got caught on and tore up the mesh seat is probably not going to fly. Small higher-res monitors are also more expensive than huge TVs, and as or more delicate. The staff would spend more time than is probably worth it telling gran and gramps that they can’t use the “little television” like that. All of this stuff has to be handled with smoothness and grace 24/7 by a desk staff that don’t regularly use these items in their professional lives. You can’t just say “it’s a computer monitor gramps don’t use it” and hang up the phone. Many people also consider office equipment ugly, and how the room visually hits when you walk in is a huge consideration. Some weary overworked travel-worn office drone would probably want to jump out the window if they opened the door to their safe place of respite only to see a the better part of a corporate workstation looking back at them.

Designing experiences can be complicated and difficult, and that’s even more true because many of the most important aspects of it aren’t even consciously perceived by the intended audience. They all just fit organically unto a unified experience.

The best solution to all these problems is to have a extra 'co working' room that any guest, or for a fee anyone, can use and just bill an extra $15/day to use it (or whatever) including the coffee machine.

Working and sleeping in the same room is actually not that great for you most the time.

Most hotels that I’ve stayed in recently have one that they call a “business center” or similar. They’re the new hotel gym.

Most that I've stayed in don't but often have pools and gyms. I suspect it comes down to if they're targeting a 'business traveler', for example it's definitely a thing in 'corporate' hotels or ones by a airport.

Most of the hotels I stay at are on the east coast, so that might make a difference because of how much business travel there is there, but even the more family-focused ones in touristy areas have them.

I have very rarely used a business center in literally years of days traveling on business. Maybe I've printed a few pages over time.

I’m pretty sure they either fall under “useless features that sound sort of useful when you’re booking a room,” or “alleviating your guilt for not getting a cheaper hotel room on someone else’s dime.”

Probably mostly the former. It's cheap to provide in a larger hotel and some business travelers sort of expect it even if they basically never use.

> There was a LTT video where they mentioned how WiFi used to be better in some locations so those rooms were more desirable and the hotel's solution was to make it standard for everyone. They seemed to be suggesting that they brought down the quality rather than balanced.

This was from their podcast 'WAN Show' a week or two back, specifically about hotels in Las Vegas.

I agree with your point, generally, but after COVID-19, remote working is opening new use cases: I occasionally like to travel to somewhere nice, far away from the office, and work from there for a week, because I'm now allowed to.

So I too now care about a decent chair, desk and maybe even a tv I could turn into a second screen. Wifi can be there or not, I bring my own connectivity just to play it safe, this is now quite cheap. Bonus if the place is a couple time zones away from my office so I have my mornings or afternoons free.

I'm not a huge fan of AirBnB but it's been more reliable than hotels for a few of these factors: hotel TVs are locked-down and many won't accept an HDMI input, assuming there's a socket at all. Normally you're not offered (barring extravagant prices) more than just a bedroom, so the chance of table and chair being any good (or existing) are not so good), etc.

As far as business investments go, I’d need to see some really solid market research showing enough people were willing to choose a hotel for these amenities in-room rather than a one with a “business center” as many currently have, a coworking space, coffee shop, or even a public library. People wanting specialized private spaces like that don’t generally look to hotels to meet their needs, and considering how quickly hotel room outfitting expenses scale, it really has to be worthwhile. For example, an in-room stationary bike would probably be cheaper and more popular than a good office chair and monitor, but it just doesn’t make sense because enough people will be satisfied with an in-hotel fitness center. I think its really easy to assume our use cases are far more universal than they are.

As someone who has thought about this remote style, but hasn’t done it, I don’t think I’d want to be in the hotel room much. I’d much rather find a coffee shop to work from where I can get some of the vibe of the city while still working.

Otherwise, you’d only get a few hours per day in the evening of experiencing anything you couldn’t do at home, so what’d be the point of spending home rent +hotel +travel for the week?

I suppose it depends somewhat on why you're there and how well you can work in an ambient social environment. Mind you, I don't really disagree but, if I'm focused on a computer screen, I'm not sure how schlepping my laptop to a random coffeeshop is that different from being in my hotel room.

Take a look at CitizenM. I travel a lot for work and that is my go-to place due to how tailored it is for also getting work done.

I like CitizenM, the only negative is a lack of kettle in the room.

In my experience Holiday Inn Express usually has a good computer chair too.

I travel a lot and I agree that most hotels suck for work. But the primary function of a hotel is to provide a safe place to sleep. If I want to do work, my best bet is to stay late in the office where I spent the day, since most people will leave at some point. I get good chairs there, large rooms and monitors if I need. It's much better than trying to retrofit a small hotel room into something else.

Yeah and even if you’re a guest of the office I can’t imagine most places would shoo you out after your business with them is concluded.

Yeah, most of the time you just need to ask ahead of time and people will accommodate.

I usually stand at the ironing board (they're typically height adjustable). Keyboard goes on board, laptop goes on trashcan which is on board (just to get the screen higher). I usually bring some paracord so I can tie the whole contraption to something heavy so it doesn't wobble. Sometimes I'll use the cord to hang a second travel monitor from some nearby art--still working on a proper harness for it.

It's fun. I occasionally get work done too.

lol, sounds janky but fun. Would like to see a picture.

I'm surprised ironing boards are so readily available.

Not really, businessmen/women usually need their stuff ironed.

And they are ironing their stuff on their own in their hotel room? Isn’t there a service for that?

Costs money (which may or may not be reimbursed) and time. Providing an ironing board and iron are cheap and, while doubtless less common than it used to be, is probably something some number of people expect. (I've never personally used them in a hotel.)

Yeah if the hotel is nice enough then usually you can have suits and shirts pressed for a fee. If they don’t have on-site “wash-and-fold” then they probably won’t press clothes either.

Sure, but if your company isn’t going to pay it, then you are going to prefer to iron it yourself. :)

Hotel laundry services of any kind are unbelievably expensive in my experience. Usually US$10-30 per item, or the local equivalent, with e.g. socks and underwear being at the low end, slacks and collared shirts at the high end. Dry cleaning, of course, is a significant premium beyond that.

I thought it was just a US thing when I was younger, but I've found it to be true even in other countries.

Not at 2am when their flight lands, and they have a meeting the next morning.

the world will come to learn about hand held steamers eventually.

For all the problems of AirBnBs and co., every single one I ever booked was better for working with a computer than every single hotel room I ever booked within 2x price.

That seems like nonsense. Hotel chains and esp business travel ones it'svery standard for hotels to have desks and a chair. Not saying it's that nice but those are way cheaper than airbnbs usually

I think companies such as WeWork or Servcorp try to fill that gap. I don't like working in the same room that I sleep in - you have to keep the room tidy for video calls, handle housekeeping, and hope that the seat won't break your back. The rent-an-office locations are usually pretty well equipped with good desks and large monitors.

When I was traveling a lot that would have seemed like way more logistics than I would have wanted to deal with--and I doubt my company would have covered. Presumably if you're traveling on business you're in the location for some other reason than working on your computer for most of the day.

I used to travel a lot for work. Hit about 180 days one year. (OK, about a month of that was vacation.) I'm not primarily a developer but I do a lot of writing. Honestly, I've never felt particularly constrained from doing that on the road but, then, although I have a nice home setup, I don't need that nice setup to work.

And I like having a king bed even if it's just me. (I do like having a desk and some sort of office chair though even that isn't really critical most of the time.)

In many hotels there are other spaces where you can work beyond your room.

Hotel rooms are simply not designed to be places to do work in. They are places for you to sleep in, so of course, the bed is the centerpiece.

I took a look at business hotels in Japan. These are hotels explicitly designed for work travel and nothing more. Small rooms, bed, shower, but not much to actually work. And it actually makes sense. If you are on a work trip, why would you want to work in your hotel room? The whole point of a work trip is to visit a work place, that's where you are going to work, not your hotel room. In fact, from my limited experience of work travel, doing more work is the last thing I want to do when I am back at the hotel, it is often an exhausting day, and there is a good chance I have to get up early the next day, so the hotel room is for relaxation and sleep.

If you really want to work in your hotel room, or do anything other than using the bed and shower for that matter, you are probably better off with "apartment hotels" and short term rentals. If available, student residence rooms can be a minimal option for working and sleeping, that's what they are designed for. Note that there are also hotels with co-working spaces.

Maybe what you want, that is essentially a short-stay student room for grownups will happen one day, but I see many obstacles in making it a "get rich quick" investment. It may not be a great hotel for those who just want to sleep (or have other kind of fun). And if you want to eat in there, you will lack the amenities an appartement offers. And if you are not alone, a co-working space may be a better option.

I was going to say what I’m about to say as a reply to the parent, but then I saw your comment mentioning Japan.

The rooms in Clayton Bay Hotel in Hiroshima absolutely has a nice proper work desk and a work chair. So if anybody here is ever in Hiroshima Japan, you now know where to stay :)

Not sure if this applies to all room types though.

Disclaimer: I’m not related to that hotel in any way other than having stayed there one night some years ago.

You are assuming that the work place you are visiting is your own company. Most of my travel has been for consulting, demos, and such things. You go to someone else's workplace to meet with customers. It is not a place you can grab a desk and do your own thing. So when you need to prep for the next meetings, your hotel room is the perfect place to do such a thing.

By the way I have also found that most hotels suck at draining water from the shower. Usually after only 5mins the bathroom is a complete mess. I suspect this is on purpose, so that guests don't use too much water.

>I might have invented a hotel chain for work stays.

That's basically what wework is. I know you can't officially sleep there. I don't know what they would do if you slept in one of the 24/7 access plans though.

Also you are vastly overestimating the amount of "work" people do in hotel rooms that are not in tech.

I worked for a mid-size tech company for many years and did a lot of travel (not as a developer). Sure I'd check email and maybe do some writing. But, while I liked having a halfway comfortable chair, I was pretty much content to work with my laptop on my lap if that were the only option. I wasn't great at focusing on writing an article for one of our in-house pubs say, but that had very little to do with work amenities and more with the fact that I traveling to attend an event etc. so I had a lot of distractions.

(For an extended trip to a single location where I was only intermittently at a customer etc., maybe I'd consider asking for a co-working space but I never did and don't think I'd have gone to the trouble.)

Hotels all have 30-day maximum stays. And they always use the cheapest bandwidth pool so internet speeds are not good. And it's shared.

Why stop at just office chairs? Why not detect all categorizable objects in every photo and rank based on that?

Could also crop just the object detection regions of each image, run those cropped images through CLIP/SigLIP, then UMAP and HDBSCAN to view a 2 or 3 dimensional latent space clustering of office chair types.. might reveal some info as to what kinds of chairs exist in what geographical regions. Could use a VLM to auto-tag each cluster given a couple images from each one. Could run PCA on the CLIP embeddings and have some sliders for each principal component.. maybe the first is chair color or size or whatever

much data = much fun

I tried to detect models of specific chairs, but it's very difficult. To train the AI model, you need many photos of the same type of chair, and on the internet, you can usually find only one stock photo of each chair.

Make synthetic dataset with reposing

There's an Internet Movie Car Database, and independently, an Internet Movie Firearms Database. Surely there are more.

I feel like they should be one database with object_type=car and object_type=firearm respectively. And then I can finally search by object_type=vacuum_cleaner and find out the wild-looking ball-shaped vacuum in that sci-fi movie whose name escapes me...

The classic orange 60s hoover ball?

https://d3j17a2r8lnfte.cloudfront.net/mvh/2024/3/medium/bzSo...

My grandparents had one in the 70s, it always amazed me.

This is pretty cool; however this account has been repeatedly posting this same story since 2023 and never anything else... weird

https://news.ycombinator.com/submitted?id=nomad86

Well, it's been a while since it was last posted, so seems fine. Interesting how it only caught on once it triggered some techy interest.

I can understand why. My thought pattern was also like "oh YOLO, I know that, interesting to see some application for it, oh cool idea, upvote".

Interesting idea and execution! So basically you wanted to find hotels where the rooms have office chairs and a desk? Or just in any of the images, e.g., a lobby?

Side note; I love how YOLO, a deep learning based model, is now being referred to as traditional object detection. Template matching gang rise up.

In our hotel photo database, we have everything: rooms, lobbies, bathrooms, pools, exterior shots of buildings, etc.

I trained an AI to recognize ergonomic chairs, but sometimes there were errors. For example, a chair in a hotel's SPA was always identified as an ergonomic chair. That's why we manually reviewed all 50k photos to verify them.

50k manually? How long did it take you?

I don't know, I assigned this task to trusted specialists from India.

Wow, manually reviewing 50k photos is a lot! Would you be willing to share what the cost of that was?

I created an app similar to Tinder that facilitates manual verification. Around 60 photos can be verified in one minute. The whole process took about a week and didn't cost much.

Sounds like YOLO verified the work of the specialists too! (;->

Doesn't sound that much. When I was playing with datasets, for simple tasks I only took around 3 seconds to classify an image. That's 1200/hour or on the order of 40 hours of work. That can't cost much when outsourced.

RLHF in the wild, nice.

> I trained an AI, but sometimes there were errors.

I edited that down to a proper summary of AI in general

This desk is pretty useless, despite having an office chair: https://www.tripoffice.com/nigeria/lekki-ng/souz-suites-apar...

> Or just in any of the images

I just looked at hotels I'm familiar with, and images with highlighted chairs were not limited to the guest rooms. Some are definitely from shared spaces at the hotel.

I've checked locally. Most of these aren't really hotels, but short term rentals.

I don't have an huge issue with short term rentals per si. They are an important niche for tourism when you take the whole family, whereas most (esp. city) hotels are not really appropriate for a family of 4 or 5.

OTOH, (esp. city) hotels are usually fine for the business trips they were designed to cater too.

This leaves us with “digital nomads.” Helping these find ways to put additional pressure on the housing market through short term rentals will only cause locals to get politicians to further restrict them.

Nice idea! What did you use for your hotel dataset source? I've been wanting to work on travel based app idea with hotel integration, but seems like most providers (like booking.com) lock down their data.

> What did you use for your hotel dataset source

Do you hear that noise? It sounds like something is scraping

perhaps, but they build out a full booking flow that doesn't take you off-site: https://www.tripoffice.com/united-states/new-york-city-new-y...

This is not scraping. Many hotel sites share photos with their partners, we work with Lexyl (HotelPlanner).

You can sign up as an affiliate with both Booking and Expedia to get API access to their data. It's meant for people who are going to run their own hotel booking sites with Booking/Expedia content, so it's not quite as trivial as a random free signup, but it shouldn't be too hard to do for a real business. OP's site appears to be affiliated with Lexyl, which owns some other hotel booking sites, so I assume they already have this access.

That said, I would consider scraping, even with API access. In some ways the API access is both limited and binds you to their terms of service, and depending on the legalities in your jurisdiction, scraping could be more effective.

Another use of YOLO, beating CAPTCHA puzzles: https://arxiv.org/abs/2409.08831

How did you go with duplicate detection? (For context, aside from being amazingly generally useful to anyone with non trivially edited images, lots of hotel chains reuse content)

Perceptual hashes are very good for that, maybe with some adjustments for mirrored images and some crops

Yes. Most of the systems I used a couple of years ago (immich for instance) were still quite immature.

Can use dhash that is more robusts to compression, crops and color changes. Mainly look at the features in the image.

The README explicitly says it's not good for crops, and from the algorithm description it can't be. Or is there another thing called dhash with a different algorithm? https://github.com/benhoyt/dhash. https://github.com/Rayraegah/dhash is similar but horizontal only.

I have no programming expertise but I'm looking for ideas that use YOLO to detect stuff like this. I was thinking of a way of automatically tagging and collecting data from my daily commute dashcam/GoPro footage - number of EVs, demographics of traffic, etc.

I want this to be a project that teaches me the ropes but since I need instant gratification, I'd like if the result also offered value to others.

The one thing I want to avoid is cleaning up data, since spreadsheets give me the ick.

Have a look into Sam, groundingdino, groundedsam

https://github.com/IDEA-Research/Grounded-Segment-Anything

You can use that to take images and generate annotated segmented images/masks that you can then train a YOLO model on. I've done this for prototypes before and it's a very quick way of getting started as you can hand off the really annoying annotation work to a machine.

Awesome! How long did it take to run locally on your Mac? Any details you can share about the stack used for processing (other than YOLO for detection)?

Thx! Downloading all the photos took the most time. Object detection process took only several days.

At TripOffice, we use simple and widely-used tools: Python, NextJS, and MySQL.

Even shared workspaces often get this wrong. On their pictures i see desks with crappy chairs and no big monitors to connect my laptop to. Bizarre.

What I'm most curious about is how does one get 40 million hotel photos? Is this offered through some Booking.com API etc.?

I was also wondering this. The hardest part about this would be scraping 40m photos.

All major booking sites share hotel photos with their partners.

We are in several partner programs, but now we mainly work with HotelPlanner.

From a few minutes of clicking around, most of these "office chairs" look entirely unsuitable for longer periods of work. I've also never seen a hotel room without at least a basic "school chair" and those are far more comfortable than a lot of the plastic slab or designer clamshell things I see in these photos.

Hey, cool idea. Would you be able to tell me the tech stack for the whole app? I want to build a similar application for some other use case. I have built a static map with all my labels using leaflet in Python. To turn it into something like you have, what technologies will I need?

Thx, see here: https://docs.mapbox.com/mapbox-gl-js/example/cluster/

I use MapBox, but similar things can also be done on Google Maps.

This is really interesting. I’ve been looking for a way to automate the next step in this process after you know which photos have chairs: knowing the model and manufacturer of each chair. Unfortunately training a model suitable for the long tail of possibilities is beyond me.

Ok, I'll bite: why have you been looking for ways to identify the make and model of hotel chairs?

I run officesnapshots.com and we currently identify visible products manually (not just office chairs).

This allows us to show our readers which specific products are in the photo they are looking at which is a service we offer to the manufacturers.

It is advertising, but like hyper-specific and relevant to the photo you’re looking at.

This is a great idea. What I have found worse in hotel rooms is not the chair, it is that the desk is at a stupid height relative to the chair. I use a standing desk at home and would love an easy way to use a standing desk at a hotel without carting around a tripod.

I travel extensively for work. Where are you guys staying that doesn’t have a basic desk and chair with outlets and stuff ready to go? Even the most basic Marriott property has a totally comfortable workspace in every room

I travel extensively for work, and in Europe this is a real Problem. I stayed in Steigenberger Hotels in a "Business Suite" that does not had a chair and a desk. Marriot properties like moxy for instance are guaranteed to have no desk in the room.

I stayed at the Hyatt Centric in Boston. Take a look at the photos - no desk or chair.

Moxy (Marriott) hotels tend not to have desks either.

What can YOLO detect actually? I am interested in satellite imagery, what are the options in this field?

Disclaimer: i am working for myself, i have no money, a 15 year old laptop and obsessions

YOLO is a pretty simple and flexible architecture. Like most of these models, you can label some data and just freeze the intermediate layers and just retrain the final predictor head.

Checkout as well darknet, which runs at really high fps on super cheap hardware

You should remove pictures that contain multiple office chairs, i saw a lot of pictures that got properties marked as postive beeing their business centers or conference rooms.

I would like to do a variation on this theme where instead I scan pub photos for pool tables and create an index in London.

It can be hard to find a pub with a pool table these days!

Heh. Other than a place in SF that had a bunch of pool tables near the Moscone, haven't seen one in a bar in ages. There used to be one in the bar that we hung out at in my grad school.

You posted this in 2023 too. Thought it sounded familiar!

All 8 of their submissions are versions of it: https://news.ycombinator.com/submitted?id=nomad86

Some hotels I clicked on are being recognized for their 'business centers', that one crappy little room with two ancient desktops and a printer.

There is a project that helps the government identify locations in child sex abuse photos, I wonder if a project like this could help.

Unfortunately digital nomads aren't going to be staying in hotels. Maximum 30 days, plus terrible internet.

https://thewirednomad.com is better suited because they are all non-hotels and have verified internet speeds.

Previously posted on HN: https://news.ycombinator.com/item?id=37176439

How do you get access to 40M Hotel Photos?

Do you have any insights on which hotel brands have the best chairs? Is furniture standardized across brands?

I haven't checked it, but such data should be easy to get from our database.

How happy are you with the results?The execution is great, but I feel the idea didn't really validate well. Every single sample ended up with pictures like this: (https://hotel.trvcdn.com/380/de/800024805/700034633/nh-colle...). While I'm impressed at how accurately you can detect chairs with wheels, that doesn't seem to be an indicator of an ergonomic workspace. Quite frankly, I would have preferred (and been fine with) sitting on the bed with my laptop in every hotel I clicked on.

I wonder how good Gpt and Claude could do it. What would be the computational cost difference?

At first I wanted to use Google's Vertex AI for this purpose, but the costs were too high.

Besides, in this case, we would also have to upload 40 million photos to the cloud for Google to evaluate what’s on them.

YOLO is the best for such tasks; it works locally and is really fast.

Did you do anything to retrain the model, or just use it out of the box?

We manually labeled almost 1,000 chairs in various photos to train the model.

How much did you end up spending?

They did it on their own computer. https://github.com/ultralytics/ultralytics

Without at least some technical details and insights from process of doing so this is just ad for your website. And it does not even work in country with 146 million people, but author is from "democratic" european country so not surprising.

Reply on news.ycombinator.com