Datasets:

Bakanayatsu
/

illustrious2.0-character-regularization-overkill

Tasks:

Modalities:

Formats:

Languages:

Size:

Libraries:

Update README.md

1937434 verified 3 months ago

618 Bytes

task_categories:
  - image-to-image
language:
  - en
size_categories:
  - n<1K

Features:

High quality images (Removed anatomically incorrect bodies and hands)
Varied resolutions ("768x1344", "832x1216", "896x1152", "1024x1024", "1152x896", "1216x832", "1344x768")
Quality classifier code/workflow using SmilingWolf/wd-eva02-large-tagger-v3 as feature extractor (WIP)
Filenames refer to the number line in prompts.txt

Used for:

Dreambooth/Full-finetuning regularization images for Illustrious v2.0 Stable
Lora regularization (Regularization images must be lower than 1/4 of total repeated dataset images)