Comment by DarkShikari | Hacker News Reader

Comment by DarkShikari | Hacker News Reader

Parent Full thread

Whether there's good money in solving this sort of problem is another question -- but sorting through shared photos at events is a surprisingly hard problem, as I learned recently.

I attended Anime Expo 2011 in costume. Over the course of the two days I was there, I probably had at least 100 people stop me for photos -- not unexpected with thousands of people walking around carrying cameras. I attended a photo shoot and the whole deal: surely it'd be easy to find at least one of these pictures later, right?

After the event, some friends were curious and wanted to see a photo of me. So I just search through the albums uploaded to Flickr and...

... well, it took days after the event for anyone I knew to find a single photo that even had me in the frame. The volume of uploaded photos was so high that you'd probably have to search for hours to find anything in particular. Nothing was tagged. Want to show your friend a really cool cosplay you saw? Unless you took a picture of it yourself, with your own camera, the odds of being able to find it are practically zero, even if hundreds of people took pictures of it.

Now I understand the need of so many people to bring cameras with them: it's not necessarily that they need to have their own personal pictures of the things they liked. It's that if they don't take them, they'll never be able to find anyone else's pictures either. We're overwhelmed by such a sheer volume of photos and left unable to find what we want.

The problem you've described is very solvable. Photos contain the time they were taking and, increasingly, the location they were taken. If you're looking for photos of something you know you saw, a reasonably accurate guess of time and location should narrow the choices to a manageable number.

A two day period is more difficult, but sorting photos by popularity will let people find most of what they're looking for. Photos of themselves are harder, but the facial recognition technology to do so exists as long as you're not in too elaborate of a costume.

I've almost convinced myself to work on this at this point. I hate taking pictures, but I love looking at pictures of memories I have. I don't think I'm alone.

This is true for what it's worth (and you acknowledge that there may not be 'good money' in solving this problem).

If you _really_ wanted a picture of yourself from the expo, you'd have gotten someone to take a photo of you. So your interest (given that it comes from after the fact friend-curiosity) is right at that marginal level.

If you just wanted some non-specific photos of the expo without any particular goal in mind, that's already easily done, as you saw.

So what you're looking at here is people with specific goals in analyzing a photo stream but who didn't, for one reason or another, turn up and take photos to meet these goals. I don't know if you can make any kind of money - good or otherwise - from providing a service to such a marginal set of customers.

The exception might be law enforcement - find some better photos of the guy who followed a woman into a bathroom at exactly 12:42, or the person who grabbed someone's purse, etc.

Sounds like a good case for NFC. Put RFIDs in the attendee badges, and a sufficiently-advanced camera app could auto-tag those people as the picture is taken.

Does that sort of technology work directionally? Typically the place was so crowded that you'd often have 40 other people within 15 feet when taking a photo, so it'd have to identify exactly the people in front of the camera, and not to your sides or behind you.

GP conflated RFID and NFC a bit, at least their colloquially accepted meanings. What we normally call 'NFC' has a range of <20cm, often <<20cm (http://en.wikipedia.org/wiki/Near_field_communication). This is what also makes NFC more suitable for mobile payments than RFID.

Hmm; true. Maybe just passes with large, simple QR codes on the front of them, then? Then the device taking the picture wouldn't even have to decode them just-in-time; the data would be encapsulated in the picture itself.

QR code on a name tag or conference badge would be a great idea.

RFID refers to a family of technologies that operate over a variety of frequencies and given by the ISO standards. Specifically, 134kHz, 13.56MHz, ~915MHz, and 2.4GHz. The 13.56MHz standardized frequency for RFID is the same as that for NFC. The RFID ISO standards that specify the use of 13.56MHz are ISO 14443 and ISO 15693. NFC is specified in 2 ISO standards, but the more relevant to the discussion is ISO 18092 and it is derived from ISO 14443. So, there is some interoperability between ISO 14443 and ISO 18092, and therefore there is good reason to "conflate" NFC and RFID.

EDIT: I should cleanup and redact the Product Requirements Document I wrote 5 years ago for a transceiver board we built that could talk ISO 14443 and ISO 18092 (as well as ISO 15693) using a single transceiver ASIC and antenna tuned for 13.56MHz.

If you have a seed image then Google Search by Images solves this to some extent.

http://www.google.com/insidesearch/searchbyimage.html

Also I should add, you might want to take pictures of yourself or your friends in the foreground. That's what personalizes a photo for MOST people. imho.

I thought of that as well, but surprisingly, I only got about half a dozen requests to take pictures with people out of the ~hundred total. Going by what I saw of other people being photographed, this was probably typical.

The idea of Color was using data harvested from your phone to take a stab at figuring out which pictures were relevant to you.

Reply on news.ycombinator.com