olivierlacan's profile

37 Messages


570 Points

Sat, Jul 27, 2019 7:50 PM

Display high certainty computer vision tags in keywords

Right now the only way to find out what computer vision (CV) tags were detected in an image is to search for something like “person” + “smile” and hope the search engine has appropriately categorized photos.

It’s impossible to tell whether a synced image has been processed with computer vision. Understandably this may not be an instantaneous process since it’s quite an intensive process. I doubt Adobe scans images on the fly, so internally the Adobe servers must have a list of high certainty computer vision tags that are used as indices for the search engine.

I believe it would be quite beneficial for user discovery of the capabilities of computer vision or descriptive text-based image search to help users learn what the search engine sees in each image. Otherwise it’s very difficult to know what the search engine is capable of and what kind of keywords it can pull out of images.

Right now the process looks more like this:
- type a keyword like “ponytail”
- see the “Oh no! A black hole!” Error (which is still displayed sometimes before results are received)
- erase the keyword and try something else like “head” + “back”
- bis repetita



3.7K Messages


61.6K Points

2 y ago

Lightroom should not only display these keywords, it should also give you the opportunity to correct them if they are wrong. For example: I have about 100 images of elephants in Lightroom Mobile. Lightroom finds most of them if I search on 'elephant', so that's good. But it also finds an image of two rhinos and there is no way I can tell Lightroom that's wrong. Yes, I can add the keyword 'rhino' manually and I did that, but that still results in this image being found if I search for 'elephant'. Apparently Lightroom now has both 'elephant' and 'rhino' as tags...