Hybrid Dog wrote:The link in helpfiles/about.html is broken:
https://project.king-arthur.eu/projects ... st-skin-db should be
http://project.king-arthur.eu/projects/minetest-skin-db, else firefox tells it's not secure.
You could automatically detect duplicates:
When a skin was uploaded, use pngtopnm to get a pnm image file, then use md5sum on the file to get a hash value. You can then compare the hash value with the values of the existing skins.
Converting the image from png to pnm is needed to detect duplicates which have a different compression ratio.
Thanks for the report and suggestions.
I reported the expired certificate, it will be fixed (some day).
bPlaced.net is a free php webhoster, that does not allow to install additional software. I played around with
https://github.com/travail/php-Image-PerceptualHash but the result is *meh* :
http://minetest.fensta.bplaced.net/finddupes.phpSome are actually duplicates (from human view and from algorithms view) e.g.
http://minetest.fensta.bplaced.net/#id=297,165Some have just a different eye color, but they look the same for the algorithm e.g.
http://minetest.fensta.bplaced.net/#id=631,85but sadly there are a lot of false positives e.g.
http://minetest.fensta.bplaced.net/#id=1309,1307The best would probably be a AI that sees the difference of the eyecolor, but does not see the fragments caused by compression. The big companies may have one or working on one, but they probably wont share it for free. :-/
as a mid term solution I think ill implement a questionbox to ask the uploader if he really want to upload a potential duplicate.