P: Distraction Removal: People (CR & LrClassic)

When trying out this new feature it eliminated only one person when there were 3 to remove, and added a new distored figure in the photo.

I just removed some people in the background from a photo. However in one cases, the person got replaced with a different, ai generated person. Also a person was partly behind me , and it was correctly recognized, but replaced with just a shoulder and an arm. 

Needs ability for user to select and area to look at for distractions. Otherwise, removes the easiest which usually happens to be the subject of the image. And does nothing for the distracting people and no way for the user to "point" it to those people. Works quite well about 2/3rds of the time. The other 1/3rd can be humerous. I'd rate it an initial prototype, well before a beta release to users.

I'm a travel & church photographer and just finished 3 months of shooting in Italy, Slovenia and Croatia.  I have spent the last 2 weeks in post in LR editing the shots on my powerful desktop because the AI featrures of Adobe are best used on a powerful PC, although I do edit on the rtoad a lot.  Anyway, I have been using this people reomve AI feature on about 300 images and it works like magic when it works, and when it soesn't you just delete that pin.  Here are some more comments.  I have posted once earlier.

It works for me about half the time and it is wonderful.  When it does not work it is because there is something the person was standing in front of that can't be regenerated accurately.  Example, - a man is standing in front of and cutting off the view of a statue and the part of the statue that is regenerated can't be accurate.  But people sitting in pews?  It deletes them and regenerates perfect rows of empty pews.  

- We need the ability to click on a person in the scene that has not been picked up by the scan and for that to redirect the dtection effort.

- We need an option for the selection to be individual people vs a large group of people.  Why?  Because sometimes in the selected group several people are deleted perfectly but one person messes up the whole selection because whatr was behind that one persone was not regenerated anywhere near OK.  We need the ability to leave that person that the dlete didn't work well on in the scene and delete the rest that were deleted properly in the group delete.   

- I am a high-resolution Medium Format photographer (Fuji GFX) and I zoom in on every deletion to see the effect.  When it works it is perfect - same resolution and no noise added and a deletion that is completely undetectable even at full resolution.    

Used it in front of a monument to remove a tourist group standing directly in front.  It seemed to identify everyone well but the end result it left a few people in place with part of the head missing and a dog.  It's new technology which I'm excited about but surprised to see some of the results.

felicidades por la nueva herramienta de eliminacion de distractores. Para mejorarla se podria añadir la opcion de aumentar manualmente  el área que se ha generado automaticamente con alguna herramienta de pincel para añadir personas que no se han detectado automaticamente. 

gracias

Fun addition, found people I didn't even see; however, it failed to see ones I DID see.....Also, when I wanted to isolate 2 people overlapping, it couldn't distinguish the two, even tho one was facing the camera in the foreground.

I tried removing people from the Occulus, and the results are terrible. It located people on the bottom floor as well as some on the 2nd floor. The people on the 1st floor that were partially removed currently look like aliens...some are blobs, some are missings parts & pieces, It did not detect anyone on the steps or upper level.

On some of my images this works very well and on others it will pick up people in the background and miss obvious people that are much bigger in the forground. Looks like this needs work. 

Works well enough! I would love to see, however, a better general incorporation of AI removal with Noise Reduction. Currently, there isn't a lot of cohesiveness when using both tools. For example, a section removed using AI has a different noise composition compared to the AI noise-reduced image, even when the AI parameters are updated.