Today’s Question: Would it be worthwhile to upload the scans and derivative files to Lightroom (cloud) to be able to take advantage of its AI search capabilities based on subject to weed out duplicates? Or would this end up creating confusion between the two Lightroom versions?
Tim’s Quick Answer: While the AI (artificial intelligence) search feature in the cloud-based version of Lightroom is impressive, it isn’t perfect for tracking down duplicate photos. In addition, a workflow that involves both Lightroom Classic and the cloud-based version of Lightroom can be somewhat problematic.
More Detail: The cloud-based version of Lightroom includes an AI-based search feature that makes it possible to locate images based on the subject matter of photos even if you had never added keywords or otherwise identified the subjects that appear in your images. While this can be a helpful feature when attempting to locate an image, I don’t consider it a great feature when trying to locate duplicate images.
To begin with, in my experience the AI search results are never all-inclusive when searching for a given subject. Based on variations in camera angle, lighting conditions, and more, many images that contain a subject you’re searching for will be excluded from search results.
In addition, I’ve found that there are a very large number of false positive results in search results. Sometimes the search results are at least related, such as when images of helicopters are included in results when you search for “airplane”. However, other times the results are completely off, such as when a rock formation is somehow interpreted as being a person in the photo.
Because this AI search is cloud-based and is being updated based on an ever-growing library of photos being evaluated, the results will certainly improve over time. However, in my experience the results are not so good as to be very helpful for locating duplicate images.
Using the cloud-based version of Lightroom to identify duplicates from a Lightroom Classic catalog would also be problematic, since the approach to image storage for these two versions of Lightroom are so different. If you located duplicate images using Lightroom it would be somewhat cumbersome to then delete the original images from a Lightroom Classic catalog. Adding to the challenge, before you could use the AI search in the cloud-based version of Lightroom you would need to synchronize the source photos from Lightroom Classic, which could require considerable time depending on how many images are in your catalog and how fast your internet connection is.
I generally find it more effective to track down duplicate images in Lightroom Classic based on capture time or other metadata values, rather than trying to use the AI image search for this purpose.