Can you create an advanced dreambooth video for classification.
I know the video with ohwx man.
I mean a new one where it learns "jumping man mid air towards a hanging rope on the moon" kind of thing.
So how it not only learns a new dataset like a face and adapt the rest it, but also a whole new situation where the rest of dataset can adapt to - while learning multiple new keywords at the same time(is that even possible?).
And the last thing is classification images - if I have a "jumping man towards a hanging rope" - image(like indiana jones) can I create a textfile with the same name as the image, with the keywords for that image? if yes, how do i tell sd to use that textfile - and many other files too(maybe even classification by folder structure? /jumping/man/rope/indianajones001.jpg).
Can I only train the same keywords, so like only "jumping man towards a hanging rope" or can I also mix it with images of only "jumping man", "rope", "jumping man on on the moon"...
I hope the explanation is clear enough. Because I was struggling to find pictures for a situation and in detail classify it (do I use ohwx jumping man, or use the full description for that photo).
To know and understand this would boost my ability to use sd 100x better
