I trained a model on 339 pictures of monsters in a box. Since I want to understand how LoRAs work, I realized that the quality of the images has a fundamental effect on the results. I plan to remove the lower-quality images one by one and replace them with better ones. It would be great if you could support me in this process!