P: AI Culling

I am so happy that this has fruition. It's a start and work in procgress. I look ofrward to having a run at it this weekend and see how it will work with culling a wedding (yes, I know it's mainly portrits and headshots at the moment). But none the less I'll give it a try.
Thanks,

S. 

Another thought; I often bracket exposures and then use the best one or combine two or more if the histogram shows black or white cluipping. AI could select these HDR groups and do the merging for me. Similarly it could detect frames with ovelapping content and merge the panorama for me. Although it may get confused by similarly composed images that were not intended to be panoramas.

For nature/wildlife photographers who typically shoot thousands of burst frames in a day it wouldn't seem practical to have to upload all these to the cloud in a synced LRc collection to have them culled by AI. This needs to happen locally.
My criteria for culling would be Subject in focus or large part of the image in focus, Exposure within the histogram at both ends either for the whole frame or the subject. The more I think about it the harder it gets to come up with fool proof (AI proof?) criteria.
The user would need to select culling criteria and adjust strictness.
The user would also idealy have a choice of marking "good" and "bad" images by flag/reject, rating or colour label. Maybe even different ways to mark images that are good or bad for different reasons. I would probably still have a quick look through the rejects in case some excellent images had been missed. For some projects an adequate number of good enough images may be OK without searching the rejects for more.

I'd STRONGLY prefer that they fix bugs and the broken UI rather than keep adding on features. Most of us don't need vertical solutions like AI culling and would rather that was left to third-party providers.

One problem with feature requests is that niche features take effort away from other development that would benefit a large number of customers.

How about just a simple "out-of focus" tool?  I shoot airshows and will come home with 50,000 images from a week's travel.  The number one reason I reject an image is it's out of focus.  Would be awesome to have a tool that will go through them and reject-flag those that are OOF.  I mean a LOT OOF (I've found Topaz can salvage some pretty blurry images)

I shoot volume shoots as well, on commerical shoots, A/E/C, lifestyle, adverts, etc, in a documentary way. Lots of culling, 2,000 to 3000 cpatures down to appx 400 - 600 deliverables.
Here are some suggestions for feature addition in LrC. I suggest you make this happen if you intend to justify this recent doubling of subscription costs...
The catch in culling is that it is somewhat subjective based on composition preferences. So, I don't need AI to make art for me, I much as I need it to do my dishes...
That means looking at metadata time stamps, exposure, focal lenght, etc to determine clusters of images.
Then evaluate for "technical flaws"
Remove shots with obvious issues: major clipping where exposure was adjusted in successive shots, motion blur, S.S. to focal length camera blur, etc
Look closer at images, evaluate where subject appears in focus, Eye Detect AF got the correct (front) eye in focus, etc.
Add a final pass (optional for user to select) that takes a more nuanced, object reconiition driven approach, to help on more subjective selects.
Use Red label for first pass selects, 1 star for next pass selects, 2 star, for next pass and so on.
Make the culling directable by user defined objectives, so different shoots (wedding, corporate event, sports, etc) can define criteria specific to their genre of photography.

gostaria de sugerir que o lightroom pc tambem tivesse a opção de selecionar as melhores fotos com IA assim como nas versoes mobile! 

This. 

I use Narrative Select for this - another thing to buy, but it loads previews instantly (like photomechanic) and rates things between 3 quality categories, zooms into faces, etc - and now even provides LR edits, although I'm not a fan of the editing.

Have you seen this sneak peek related to this feature request? https://www.instagram.com/reel/DLBHdBPNV-N/?igsh=azJwdmsyNmw2MWpv