I just commit a change
I have found why, the mask does not work…a simple forgetfulness on my part !
Now mask works (I hope)
If blend is at 0, the system works, it take into account the mask to improve shape detection, without other change on the image
if blend < 0 (substract) or blend > 0 (add), you mixed the “mask” (the difference) with the original, but before workind on it with locallab.
Of course transition works.
I hope no error, and no problem with memory (of course this functionnality use a lot of memory!!)
Works very well indeed. Easy to use and with a bit of tweaking can modify complex shapes easily. I find the ‘Show modifications without mask’ very useful, in combination with the ‘scope’ setting, so you can see what areas it will act on.
@jdc
locallab woks fine, many thanks to all for developing this very nice tool!
By using the local-tab the hand symbolt changes to local symbol (crosshair with arrows) and everything works fine. After hitting ‘h’ or clicking on that symbol it changes to hand symbol and the local indicators are gone. I found no way to get them back other than changing the tab and reactivating the local tab. Is there an other way to get controls back, or schould there be an other way?
If the actual version proves to be stable perhaps a ‘preliminary endpoint’ in the development could be defined to get it merged into dev? I would greatly appreciate to see it merged in dev.
Looks really good! I started using the locallab branch just 10 days ago - it allows far better corrections when for example a red poppy or a red car insist on being orange
All works very well. This is a very useful too which greatly adds to RawTherapee’s capabilities. For the majority of cases I find it much quicker than drawing masks in other programs. Are there plans to merge it with the main version?
Thank you very much.
Effectivly in most cases, RT_spot is much quicker than drawing mask…
In some cases, foliages… this technique is clearly more efficient.
I put a lot of time to find the 2 key algorithms
general case with deltaE
structure detection with “Sobel Canny + guide filter” : we can certainly improve this the second part