Another victory
for automated intelligence in photography

A few weeks back I wrote about my success in working with ChatGPT to manage some very complex text, making it editable, and making it possible for me to re-publish two out-of-print catalogs of matrices for Linotype and Intertype machines. It was an extraordinary success for me. You can read that post here.

Since then I have been experimenting with the AI systems offered by Google and Adobe, and hoping that these offer some assistance to me when working with high resolution images.

The first of these was a routine clean-up of a photo where a giant traffic signal was in the frame. I wanted it gone. I did it the old fashioned way first, using the lasso, content-aware fill, the clone tool, and paintbrushes. It took me the better part of an hour to complete, and it was perfectly acceptable.

When I showed this image to my friend Jason, he said, “I can do that with AI in 30 seconds.” I accepted his offer to demonstrate, and we arranged a Zoom meeting where he published his screen and demonstrated the process. In the end, including teaching time, it took about ten minutes, but the AI Erase function in the beta version of Photoshop (in partnership with Adobe Camera Raw) did the job much better than I had, and it accomplished it in a just minutes.

Following are the steps that he showed me:

This is the original image of San Luis Diagnostic Center with the traffic signals, a truck, and various shadows. These images were captured in Adobe Camera Raw. I have selected the traffic light on the right edge using the Eraser tool. Once that selection is made, I hit Apply, and the signal was removed from the image.
My friend Jason cautioned me not to select too much for the AI engine to work on, as that will often cause it to fail. Instead, I have selected a street light at the top of the pole. This was easy for the AI erase tool to remove.
Here, I have selected the horizontal beam of the signal with its many lights. Curiously, the AI erase tool had no difficulty removing the signal and repairing the tree.
In this image I have selected more of the arm of the signal. It passes over the terra cotta tile roof and into the stucco exterior. I hedged, and selected the part on the end.
Selecting the large vertical pole and its signals was most challenging for the AI Erase tool. It had to remove the steel post, then patch the stucco wall, the planter box and, most importantly, the tiles at the soffit. The tool took only about a minute to do this, and it did it flawlessly. It even replaced the right-hand planter on the white wall.
Here I have selected the street sign. The AI Erase tool removed it easily, I left the brackets unselected; I would get those with the clone tool.
Removing the third traffic light was easy.
The shadow left behind of the pole was removed.
Tire tracks on the pavement were removed in a group.
Last, the truck on the right was removed. I had to do this twice, as it didn’t work correctly the first time.

Elapsed time? It took longer – much longer – to write and illustrate this post than to make the corrections on the photo. 11 minutes and 20 seconds.

What have I learned? Like Photoshop’s Content-Aware Fill tool, with which we are all familiar, it’s smart not to bite off too much in each action. It would be impossible to remove all of the elements of the traffic signals in one pass. It works better in small morsels.

Overall, the photo is much better after. The clutter of the traffic signals, the tire marks, the sign and various artifacts made the original too busy for me, and for the client.

About Brian Lawler

Brian Lawler is an Emeritus Professor of Graphic Communication at California Polytechnic State University, San Luis Obispo and was a Guest Professor at Hochschule München from September, 2021 to September, 2022. He writes about graphic arts processes and technologies for various industry publications, and on his blog, The Blognosticator.
This entry was posted in Art, Photography, Photoshop techniques, Software, Technology and tagged , , , , , , . Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.