Nice Trophy
<<Hills |
|
….This was Kerr's first win on the LPGA tour.
Added by Paddy Atherstone[Paddy] on 25th April 2002. |
|
|
Picture not tagged with any keywords - add some.
<<Hills |
|
….This was Kerr's first win on the LPGA tour.
Added by Paddy Atherstone[Paddy] on 25th April 2002. |
|
|
Picture not tagged with any keywords - add some.
<sigh>
Really must get the guest-pictures-from-other-galleries thing going. See http://gallery.future-i.com/celebs/pic:kerr/
Quite
How can you go about stopping duplicate postings though? Pretty difficult I would assume? btw - as another suggestion, it might be nice to show previous comments whilst I am writing a new one…
Testing binary sameness is obviously easy, but it won't spot different versions of the same pic. Arguably that's fair enough, as various people treat the same source scanned independently as conceptually different pics (personally I try and just keep the one I like best).
As for showing previous comments, I'm still trying to work out how to do it. Showing all of them will be over the top - especially on long pages. I think I may change the pic page to only show the last few comments, then have a page with all the comments and the form right there in the page. Once the feature requests site launches I'll try and et some discussion in, and give folk the options they want.
Fourier transforms. That'd be a fairly accurate way of determining a level of similarity between two pictures despite rotations or moderate transformations of the image. It would, of course, require a fair number of spare processing cycles.
I'm far from convinced with the capacity to spot 'basically the same picture, probably scanned by someone else, maybe with a different background composed in, perhaps with text removed' without a huge number of false positives.
On the other hand binary checking is easy and sufficient for most purposes I want to solve - notably dopey people whacking the same picture up more than once (I've done it once I think) and people lobbing up the same 'popular' pic independently - like this one.
As a first approximation I'll check length & md5, if they match then I'll probably do the expensive check, to be sure.
Is anything stronger actually useful?