narugo1992
commented
8 months ago i dont know what u mean slow? i think it's quite okay to have 2it/s when u using so many ai models (monochrome filtering, image classification, face detection, halfbody detection, person split, similar check, at least 6 models in your script) to clean your dataset. dont expect to have 1000it/s even when u are using this pipeline on 3090.
btw, do NOT use 3D, it is 3d. they are different, 3D is not 3d, if u use that, the 3d images will be filtered out. 😄
SaveExporter('E:\xlzl\reg\1_zumi score10\新建文件夹'): 142it [06:20, 4.19s/it] LocalSource('E:\xlzl\reg\1_zumi score__10'): 140it [06:25, 1.57s/it] Loading from 'E:\xlzl\reg\1_zumi score10': 13%|██████████▍ | 440/3433 [06:25<26:08, 1.91it/s] this‘s my code from waifuc.action import ModeConvertAction from waifuc.export import SaveExporter from waifuc.source import LocalSource from waifuc.action import NoMonochromeAction from waifuc.action import ClassFilterAction from waifuc.action import FaceCountAction from waifuc.action import ThreeStageSplitAction, FilterSimilarAction
if name == 'main': source = LocalSource(r'E:\xlzl\reg\1_zumi score__10') source = source.attach( ModeConvertAction(mode='RGB', force_background='white'),NoMonochromeAction(),ClassFilterAction(['illustration', 'bangumi', '3D']),FilterSimilarAction(),FaceCountAction(1), ThreeStageSplitAction(),)