I asked my friends for a sketchbook and scanned all the drawings from there. Train made on the model deliberate v3
I trained on 640x640 pictures, the dataset consists of 165 pictures and repeated each of them 100 times, resulting in 16500. I ask you to test in grids on the subject of portraits, various characters, pictures where there is not one person, etc. Because I do it on gcolab. Therefore, feedback is very important.
It is interesting to hear suggestions for improving and refining Lora.
Thank you and enjoy your creativity.