P: Assisted Culling Feedback Thread (LrC)

Wildlife and birds’ pictures are often shot with burst of up to 20 or series of consecutives shots, so the process of selection for the best focused picture is time consuming when you come back with hundreds or thousands of pictures that many of them are very similar at first glance.

Is it possible to add culling to this type of photographies in LRc? To the best of my knowledge the AI-culling feature is currently only for people or portraits pictures.

Thanks a lot.

I think someone needs to take a hard look at some basic mathematical practicalities in this automated culling system.

Ruling out focus or exposure issues helps a little but those can often be quickly identified by a regular human quickly glancing at the thumbnail gallery so the benefit of this assistance is limited.

Face detection is good to have but analyzing those faces just for eye focus or having them open hardly accomplishes anything in the way of culling since most cameras will auto-focus on the eyes by default and photographers employing manual focus will strive to do the same. Capturing someone mid-blink is statistically rare and therefore most photos of people with their eyes closed are quite likely to be intentional so this hardly accomplishes anything in way of assisting the culling process.

For journalism, event photography, etc, it would be more useful to detect expressive features such as smiles since you will capture thousands of mundane photos in hopes of catching just a few moments of joy, serious focus, or animated discussion, and then spend many hours searching for them in post.  There are plenty of AI models trained to recognize expressive features in images so it is puzzling not to see that capacity in a speciality photo organizing/editing suite like Lightroom.

Sorry to say but Adobe LrC is late to the game. There are other DAM & stand-alone software better suited to Ai culling. Besides, Ai is only designed for people photography not for wildlife or landscape. It also uses too much resources requiring users to buy ever more expensive computers. Too expensive for the average enthusiast user. This is meant for high volume photography businesses to help then reduce their employee payroll cost.

Since this is the feedback thread… I found AC not very useful because of its absolute, rather than relative nature. What I mean is, a shoot can have all sorts of photos, and measuring them with a single, say, subject focus yardstick rarely works. Yes, it works for *groups* of similar photos where you really need to weed out the bad takes - but it is precisely this grouping that is the recipe for it working as designed.

Imagine it compared photos *within* such groups, rather than across an album. The sensitivity slider will then be relative to that group, rather than absolute. Let’s say I have made a sequence of exposure bracketed photos. Comparing them by subject focus to weed out the one where the exposure proved insufficiently short to eliminate camera or subject movement makes perfect sense. Comparing them by subject focus with an entirely different set of photos within the same album makes no sense at all.

The easiest solution would be utilizing the same auto stacking feature to divide photos into similar groups for subsequent culling analysis.

Assisted stacking by visual similariry makes too large stacks. I expect that it would stack similar shots (taken in burst or sequence), but it stacked by topic. I was shooting a cafe on a farm, and it simply stacked all cakes together despite them looking complely different, chickens and rooters together, all different kind of flowers and plants together… I did put similarity slicer to miimum, but stacks were too large anyway.
P.S. Assisted culling worked well but very slow. Would be nice if users could cancel in the middle.

Still no answer to the question How to completely disable Assisted Culling?

I hope that Lua plugin & smart collection search term accessible for Assisted culling analysis results.

How the heck do turn off or disable Assisted Culling?

This is a waste of time. 3 people: 1 with sunglasses, 2 looking away with eyes closed = all three eyes open. And i have hundreds like that. Same with sharpness - delete good ones, keep junk. Even statistical guessing would net better success rate. Maybe ask the latest AI model to fix it!

Besides the tool being very slow. 

1. I observed it doesn’t use the full resolution image by default but some kind of preview > all the images where “out of focus” which isn’t true. 
2. why is there no button to start, pause the culling for example only on a selection? And a button to re-trigger the analysis (for example after triggering 1:1 previews) 
3. There is no setting for the open-eye-strictness. I wanted to use it to select the best pic of a group shot out of 40 images. It wasn’t possible it was rejecting 2 images - that’s it > not helpful