How image generation models are trained? Why they understand prompt words and can generate the images from texts.
In: 5
Having trained a LoRA (Low-Rank Adaptation) for Stable Diffusion, there are 2 pieces of data that I had to prepare to be fed into the system.
I have a folder of images and text files with the same name as the image, the text files are simply a bunch of keywords in a list separated by commas that describe what aspects of the images are.
The words are associated with patterns in the images and they’re used to guide the image generation algorithm towards a desired result, if you generate an image with no prompt at all, it will make a complete guess and you have no idea what will appear.
Latest Answers