On 2013-03-11 04:46, Ulrich Pegelow wrote: > The key problem as of today is an automatic detection that is way off > and which the user fails to observe. > > A few ideas: > > 1) we could go back to the strict search rules. A lens would only be > accepted from the database if there is an exact match. > > 2) we could add a second cross-check (maybe Levenshtein) of the found > lens with the EXIF string. If the two differ too much (*) we reject > the > found profile and fall back to manual selection.
For me there needs to be more feedback to the user, so that they *can* observe what's happening 1) display both the string reported by exiv2 and the string from lensfun it loosely matches 2) display the correction parameters so that they can see they're nonzero or even modify them manually (I know that's been rejected before) Strict rules wouldn't great- I have one that only matches because of the loose matching rules. If you went that way, it would be good to have a local aliasing system so you only have to educate it once about your lens. One other thing to chime in with here. Try turning lens correction module on and off when it doesn't have a match for the camera. The image doesn't noticeably change, but the histogram does! Russell ------------------------------------------------------------------------------ Symantec Endpoint Protection 12 positioned as A LEADER in The Forrester Wave(TM): Endpoint Security, Q1 2013 and "remains a good choice" in the endpoint security space. For insight on selecting the right partner to tackle endpoint security challenges, access the full report. http://p.sf.net/sfu/symantec-dev2dev _______________________________________________ Darktable-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/darktable-users
