Live data from Hacker News

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

news.ycombinator.com

31–40 of 129 posts

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

#32
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

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

#33
post #10

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…

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, w…

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.

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

#36
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.

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

#37
post #29
post #16

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.

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

#38
post #10

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…

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.

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

#39
post #29
post #16

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

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

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

#40

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?
Post reply on HN