|
Can anyone help me with creating embeddings. I want to train 3 separate embedding on 3 subjects. (or is it possible to train on all 3, that would be awesome) I don't know how to tell Automatic1111 to choose only the dress from the photo and not jewelry, choose just earrings and nothing else.
when we call dress, it should apply only dress type, and similarly with jewelries and earrings. |
Replies: 2 comments 1 reply
|
you need to describe everything except the subject in the training data. so you describe the jewelry while training the dress to omit it |
|
thank you. I found my answer. |
|
That's a really specific and interesting challenge. Training separate embeddings for dresses, earrings, and jewelry in Automatic1111 can be tricky because the model naturally tries to learn the whole image context. For your goal, training 3 separate embeddings is definitely the cleaner approach. When you're captioning your training images, be very precise, for earring images, caption everything else in the photo but clearly label the earrings, so the embedding learns to associate the prompt with just that element. Since you're working with earrings specifically, you might find some inspiration by looking at how professional artisans highlight their designs. I recently came across a website with a beautiful collection of earring crystals that really showcases the detail and craftsmanship in each piece. Seeing how they present their work could give you ideas for how to frame your training data or prompts to get better separation between subjects. |
you need to describe everything except the subject in the training data. so you describe the jewelry while training the dress to omit it