>>855
This is me, forgot which site I was on.

>>853
>>854
I want to add a pixel duplicate cache to the database and let you queue up these easy decisions differently in the dupe filter. At the moment, I calculate that on the fly when each pair is loaded, so I can't search for it, so I just have to write a maintenance task for it (and also do some ICC colour profile detection stuff to weed out this special false positive case).

The pixel-duplicate queue will be the first place where I try out automatic duplicate resolution. It will be completely optional, but you'll be able to set 'if a png and a jpeg are exactly the same, say the jpeg is better and delete the png', and it'll activate even on file import with a nice import status to reflect it.

Once we have db cache and automatic decision tech prototyped, we can explore automatic high-confidence jpeg quality comparison too.

I expect dupe improvements to do well on the 'big work' poll I'll put up for Christmas.