[Closed] when digital photographs meet statistics

Interesting.. but I'm not convinced by some of it.

He's suggesting that people are renaming images to IMG_2000 or IMG_9999.

Really? Why on earth would you do that? If you're going to the trouble of renaming it you're surely going to call it something meaningful me_and_jim_glasto.jpg or something.

Depends on his analysis.

Maybe people are renaming to IMG_2012_Jim_Glasto.JPG or people are setting the file naming rules in cameras or editing software to use the format IMG_20121130.

The date seems a logical reason for the spike in the range.

I'm stumped on img_252 though.

Interestings stuff.

*little bit of sex wee*

Never knew Benfords law was an example of logarithmic decay though.
will ponder the 0532 phenomenon. Although I reckon it's a camera manufacturer issue.

**drops all pretence of work and starts to ponder the 0532 issue**

edit: bonus points for the person who can find the most interesting (SFW) IMG_0532 on google images.

when i get home im so checking on my hard drive and see what frequency 0532 pops up!

have already done it and got a big fat zero.

SO that rules out Canon 350d and Lumix FS62 🙁

of course, another thing to consider is a sampling glitch - i.e. the mechanism he used to catalogue the file names has some kind of oddity in it a bit like the The Pentium FDIV bug.

Here's a possibility. Say a popular camera comes with a particular size of memory card which can store 531 typical images. You might fill the card, clean it out, then take one picture for uploading tho the web. That'd be 532.

Or, perhaps the numbers aren't incremental - maybe it takes a hash of the date and uses that, and the firmware default date is one that comes up with that number. However if that were the case, you'd expect the curve to bump up at that point and continue decaying from there, not return to the original curve.

My Olympus cameras (all three I've had) don't start off at 0, they make a hash of the date then append a counter to that, I think.

So for olympus users you would see a big spike at (say) P20111225001.jpg or whatever it works out as if people get given them for christmas.

However given the isolated anomoly on 532 my money's on a characteristic/bug of a popular camera firmware.

Thinks Nikons just increment till you reset them, so even if you wipe the card at 1067, the next photo is 1068.

I suppose the thing to focus on is not so much WHY 0532 but why NOT 0531 and 0533 as well....

I suppose one should search flickr for examples of 0532s that dont appear in a range 0490-0550 say. and see if there's anything particular about them?

It's got to be a flickr bug hasn't it.

Most people just dump all their photos onto a harddrive and they be getting loads of file name conflcits if cameras we consistently duplicating the 532 extension.

not necessarily, if people were uploading 50 photos from 0140 to 0190 AND 1x photo of 0532 they'd never know there was an oddity?

does it relate to picture size?
e.g. 532 x 1024 pixels.

A quick googlefight partially backs his results up. I tried a few numbers either way IMG_0532 and 533 are similar and the surrounding numbers yield results about 15% lower.

does it relate to picture size?
532 x 1024 pixels.

But why not spikes at other common resolutions?

my guess would be something to do with naming conflicts.

Here's a possibility. Say a popular camera comes with a particular size of memory card which can store 531 typical images. You might fill the card, clean it out, then take one picture for uploading tho the web. That'd be 532.

It's a good point but there's one BIIIIIIIIG flaw in it!
Even for a specific camera/card combo the number of shots stored changes each time you use it as pics of different things create files of different sizes. So the capacity if the card is not fixed.

This one beats the cranks thread!

It's a good point but there's one BIIIIIIIIG flaw in it!

There are many flaws in it, that is only one.

If you plotted the number of pictures you got on a card each time it filled up, you'd probably get a gaussian curve, which would mean you'd get a spike in the numbers AROUND one mean figure but some either side. Which is not what we see.

It also depends on people taking only one picture just after they have cleared out a card and uploading that. Unlikely - it's a pretty specific usage situation.

Just throwing ideas into the pot here.

Kind of interesting, I suppose, but he really ought to get out and play more! 😀
FWIW, when I fill up a card, I put it to one side and start a new one, keeping the full one as a backup. Memory's so damned cheap now, I bought three 8Gb cards for £15 a little while ago, and they'll be even cheaper next time. I never rename pics, I take far too many, and I'm too chuffing lazy to bother. 😆

You're all going to search for IMG_0069 now, aren't you?

I know I did...

need to get out more!

OK wild theory

just searched my windows 7 pc for 532. All the images found were resized images with one dimension of 800.

Thats because prior to upload to flickr they were resized to 800 pixels wide, by LightRoom at the original aspect ratio. The original aspect ratio was 3008 by 2000.

So maybe his search picked up another image property or some how files get renamed with a pixel dimension in

The blog has been updated with the answer.

You were all wrong.

The root cause of the issue isn't a flickr bug or some sort of camera anomaly or even the size of memory cards. It is simply that there are some very strange people in the world who do some very strange things.