View single post by Eric
 Posted: Sat Dec 30th, 2023 17:35
Eric



Joined: Wed Apr 18th, 2012
Location: United Kingdom
Posts: 4186
Status: 
Offline
I have just discovered how clever this AI can be.

Doing through some old IR photos I came across one image of Iron Bridge which I hadn’t processed because it had 4 people and some workmen’s screening on the bridge either masking the railings or behind them. It sort of spoilt the image. So I set the generative infill on Photoshop with the challenge of doing more than one fill at the same time.

Using simple path/ selection to surround each person and including some of the railing and background it took 30secs to mask and save each section in turn, then recall them en mass. I’ve focussed in on the 3 that were centre frame….




Click here to comment on this image.

I then hot the generative fill button and ….




Click here to comment on this image.


And within another 30sec the process was complete….



Click here to comment on this image.

It’s done an excellent job with the two members of the public and an ‘acceptable’ attempt at the worker who was in front of the rails compared to the other two people and he also had foliage behind as well as a more acute angle reducing the railing gaps. 

Aside from the fact it struggled to get the railing spacing right (with the worker obscuring the railings for better reference), what I found fascinating and very clever, was its ability to tackle more than one distinctly different area of the image at the same time.

It’s not perfect but getting something 90% ok/ usable in under 5 minutes compared to manual cloning which would have probably taken me an hour to do them all.

Footnote: I missed the fact the middle guy had a dog on a lead, which should have been included in the mask. :banghead:

Last edited on Sat Dec 30th, 2023 17:35 by



____________________
Eric