Building an Internet Scale Meme Search Engine • Hacker News Board

Re: Building an Internet Scale Meme Search Engine

#121

Post by oefrha » Wed, Jan 11, 2023, 5:06 PM UTC

iamflimflam1 wrote: post #89

oefrha wrote: post #24

> My preliminary speed tests were fairly slow on my Macbook. However, once I deployed the app to an actual iPhone the speed of OCR was extremely promising (possibly due to the Vision framework using the GPU). I was then able to perform extremely accurate OCR on thousands of images in no time at all, even on the budget iPhone models like the 2nd gen SE. I suppose that’s an old Intel MacBook? I’d be very surprised if t…

I think he was running it in the simulator - which won't perform anywhere near as fast on this type of thing.

The simulator runs a native build though. Might not be optimized, not sure.

Re: Building an Internet Scale Meme Search Engine

#122

Post by looki » Wed, Jan 11, 2023, 5:34 PM UTC

Very cool project! I'll try to remember it the next time I'm looking for a specific image. I noticed that repeated appearances of the search term are ranked higher, which isn't necessarily productive. Also, some kind of duplicate detection would be nice. Searching for "SpongeBob" yields many copies of the same images that mentions "SpongeBob" several times.

Re: Building an Internet Scale Meme Search Engine

#123

Post by danbruder » Wed, Jan 11, 2023, 6:30 PM UTC

this is very clever. I wonder what other use cases could leverage this approach

Re: Building an Internet Scale Meme Search Engine

#124

Post by kaladin-jasnah » Wed, Jan 11, 2023, 6:33 PM UTC

dewey wrote: post #102

leokennis wrote: post #60

Earlier quoted context omitted.

It reminds me a little of https://www.beeper.com/ . It allows you to read iMessages on Android and other platforms. To make this work, they will ship you some old iPhone to act as a server "bridge": https://twitter.com/ericmigi/status/1351934418961661959 If it works, it works. But it also speaks volumes about Apple's disregard/inexperience with exposing their stuff via the web - https://www.icloud.com/ being the prim…

I think I've seen a post from https://texts.com about that they. I don't think they ship you the iPhone though, the host it themselves.

I would imagine the only scalable way to run such a service is to run macOS virtual machines with multiple user accounts for each iMessage user.

Re: Building an Internet Scale Meme Search Engine

#125

Post by joshu » Wed, Jan 11, 2023, 6:48 PM UTC

Heh, this finds a bunch of copies of a video I made. If you are going to cache them and repost them, you probably need to have a DMCA process.

Re: Building an Internet Scale Meme Search Engine

#126

Post by vlunkr » Wed, Jan 11, 2023, 7:09 PM UTC

Bad_CRC wrote: post #107

price456987 wrote: post #103

Earlier quoted context omitted.

What's the cost of building and running a cluster of iPhones vs Mac Minis?

in the article $40 second hand, imei banned and broken screen iPhones are being used so...

There's a ton of compute power available in the form of unused phones.

Re: Building an Internet Scale Meme Search Engine

#127

Post by permo-w » Wed, Jan 11, 2023, 7:21 PM UTC

I have absolutely no experience in this area and I'm curious:

is there really no open-source text recognition software that's on-par with or close to Apple's (presumably proprietary) implementation? the article mentions Tesseract. is that the current best open-source option?

Re: Building an Internet Scale Meme Search Engine

#128

Post by paulmd » Wed, Jan 11, 2023, 7:23 PM UTC

Is there a pgsync equivalent for Oracle? Spent some time building replication from a source-of-truth to a search engine at a previous job.

Wish we could have used postgres but the tools were dictated rather than letting the requirements drive the tooling.

Re: Building an Internet Scale Meme Search Engine

#129

Post by ariehkovler » Wed, Jan 11, 2023, 7:27 PM UTC

This is mad and I love it.

Re: Building an Internet Scale Meme Search Engine

#130

Post by sankha93 » Wed, Jan 11, 2023, 7:32 PM UTC

yakubin wrote: post #9

The photo under Upgrading the iPhone OCR Service Into An OCR Cluster . In the future, data centres are going to host racks of iPhones.

It is already here to be honest. I know BrowserStack and other mobile testing platforms (at Facebook and Amazon) do host real devices, both Android and iPhones, in server farms like this. Meta wrote a blog post about it: https://engineering.fb.com/2016/07/13/android/the-mobile-dev...

At one of my previous workplaces, we discussed running the Z3 theorem prover on an iPhone cluster, because they run so much faster on A series processor than a desktop Intel machine.