AI Keywording (again)
Following up on this topic of AI keywording, would the community see a benefit in having such a feature implemented as part of CYME's 3rd party search plugin for CaptureOne? What is the ideal workflow envisioned for AI Keywording in the context of Capture One? Would it be best to let you go image by image in a selection of images and validate the proposed keywords ? Any thoughts ?
Matthieu (cyme.io)
-
Where would AI keywording get its keyword list? From a list/hierarchy that I define, or would it use its own (possibly learned from images already keyworded in my library)? And if hierarchical, would the user be able to specify the depth, i.e, 1 subcategory, 2 subcategories.
I'd definitely want to validate each image (or perhaps groups of same-subject images) at least until I'm confident that AI keywords match my expectations.
Frankly, well ahead of AI assistance in entering metadata, I'd want to see Capture One allow me to assign metadata (at least keywords. and description) to video clips, improve search (I'm on Macs), reorder the metadata display, and select which metadata fields show. Also ahead of AI keyword assistance is giving me a choice of having Capture One search as each search term letter is typed, or waiting for complete words; maybe I'm wrong, but currently it attempts to search and refine the search as each letter is typed. Mis-typing can cause unwanted delays as the program retries the letter-by-letter search as I type each new/corrected letter—at least have the program wait until a space is typed between words.
For AI, well before AI keywording (again, for me) would be AI Denoise and perhaps AI Sharpening. I think your developers did a superb job on AI masking. I suspect they could do at least equally well on other AI-based image processing.
0 -
NB. This is about the development of Cyme's search plugin for Capture One, not about the development of Capture One itself.
0 -
Matthieu Kopp, I understand some people would want to stay in CaptureOne for everything but I'd rather have the keywording done in Peakto where it's clear what Peakto suggests as new keywords and I can assign them if I think they fit.
However, to even use Peakto in the first place, has there been any update/improvement on the thumbnail size in for Capture One catalogs/sessions in Peakto? Small preview images just don't work for me.
0 -
Right away, I doubt that I'll be making much use of AI keywording. The keywords that I use are typically related to locations, buildings, building types, names of architects, etc. None of these are areas where AI has shown much promise, at least not to my knowledge.
I use Peakto regularly, but mostly for getting an overview of my Capture One catalogues and for finding images and opening them in Capture One. Peakto's conversational search hasn't really been of any use to me. When I search I search for file names, keywords and colour tags.
I too would appreciate larger preview images in Peakto.
0
Post is closed for comments.
Comments
4 comments