597
Replies
It seems to work fine for me:
All three variations were acceptable. A screenshot showing your actual brush stroke might be helpful in troubleshooting.
[This post contains formatting and embedded images that don't appear in email. View the post in your Web browser.]
@newplasticmachine, I may be missing something -- what's unsatisfactory about the first and third screenshots? They look like Rikk's results.
In general Detect Objects doesn't work well with long skinny things like poles and power lines, and I often find myself unchecking it. It results in selections that are too wide or disconnected or both, as seen in your second screenshot:
@newplasticmachine: "The third example is perfect, but done in Photoshop. The first example created some bizarre distortions."
Got it, thanks.
It seems that the Remove tools in both PS and LR are quite sensitive to the width and placement of the selection over the pole. On the left is a fair bit of distortion from my first attempt in PS, and on the right is an example from LR that looks much better:
Regarding Rikk's screenshot, I'm skeptical that the selection shown in white outline was made with Detect Objects. I've never seen it make a selection like that with a perfectly rounded bottom -- that's the outline of a brush stroke made without Detect Objects. My experience with Detect Objects on lines and poles is that it always make a too-wide selection, often in disconnected pieces, just as you're experiencing.
Rikk, can you show us the red outline of the selection you made with Detect Objects?
@efficient_dancer5CC7 you can't do much about it.
I want to make an example:
Original photo
I refreshed and I got 0 success rate upon several tries.
I open the image in Photoshop and entered instruction "Street"
Failure again, I was no specific enough so Ai "thought" that "street" could have a car (makes sense tbh)
But in Photoshop I can add more instructions to guide the AI
So I wrote "Street, no cars"
Success !
One of the 3 versions works.
The other ones are silly, I got giant walking furry creatures.
• BUT in Photoshop at least one can guide the AI so at last we are not at total mercy of what AI "thinks" will work.
One way to have higher success rate would be to have instructions.
.
[This post contains formatting and embedded images that don't appear in email. View the post in your Web browser.]
This is a great working example. The car is casting a shadow that extends down to the bottom edge, and when you don't select the shadow too, Remove generates a replacement that tries to match the remaining shadow. Using the screenshot with the generated car, here's what happens when you don't select the shadow:
When you include the shadow, the first variant has replaced the car with matching buildings, wall, yellow curb, and street and even generated a reflection for the pole:
AdChoices
