Quality of keywords generated by AI
The keywords generated by AI to help us finding images is one of the top selling points of Peakto but so far the quality of the generated AI keywords is not exactly high in my collection. Quality looks similar to what I have seen several years ago but as we know AI tools are advancing quickly.
What are you doing to improve the accuracy and what is it that we can do to help?
Some thoughts:
- Does it use available context from the image (keywords, caption, location,...)?
- Can we train it on our images to increase accuracy for instance by flagging good/bad results?
- Do you want to see examples of wrong keywords? This could even be a feature in the application.
Topic Participants
Harald Walker
Jibz
Matthieu Kopp