星级打分
平均分: NAN 参与人数: 0 我的评分: 未评
为避免大家被国内二手商坑,特转载国外免费模型
下载链接多为mega盘,国内无法访问,有需求在本帖下留言,我有空的话转存到百度盘
格式比较乱,见谅
用正常数据训练的模型SFW Model
Full Face
LIAE- res: 128, dims: 128/80/48/16, 1.000.000 iterations, Full face (SFW - old pretrain dataset)
LIAE-UD- res: 192, dims: 256/64/64/22, 1.000.000 iterations, Full face (SFW - old pretrain dataset)
DF-UD- res: 128, dims 256/64/64/22, 500.000 iterations, Full face (SFW - FFHQ pretrain dataset)
DF-UD- res: 160, dims 256/64/64/22, 500.000 iterations, Full face (SFW - FFHQ pretrain dataset)
LIAE - res: 160, dims: 256/64/64/22, 1.000.000 iterations, Full face (SFW - old pretrain dataset)
DF-UD- res: 192, dims 256/64/64/22, 637.000 iterations, Full face (SFW - FFHQ pretrain dataset)
LIAE-UD- res: 192, dims: 256/64/64/22, 500.001 iterations, Full face (SFW - FFHQ pretrain dataset)
LIAE- res: 224, dims: 256/64/64/22, 322.144 iterations, Full face (SFW - old pretrain dataset)
LIAE- res: 192, dims: 192/64/64/22, 616.730 iterations, Full face (SFW - old pretrain dataset)
DF-UD- res: 224, dims: 256/64/64/16, 1.030.808iterations, Full face (SFW - old pretrain dataset)
LIAE- res: 256, dims: 256/64/64/22, 1.000.001 iterations, Full face (SFW - old pretrain dataset)
LIAE-UD- res: 256, dims: 256/64/64/22, 1.000.151 iterations, Full face (SFW - old pretrain dataset)
DF-UD- res: 256, dims: 256/64/64/22, 400.000 iterations, Full face (SFW - old pretrain dataset)
Whole Face:
DF
- res: 128, dims: 256/64/64/22, 503.304 iterations, Whole face
(SFW - old pretrain dataset)
@
LIAE
- res: 128, dims:
128/64/64/22,
1.021.612 iterations, Whole face
(SFW - google pretrain dataset)
@
DF
- res: 224, dims 128/32/32/22, 101.000 iterations, Whole face
(SFW - old pretrain dataset)
@
LIAE-U
- res: 224
, dims: 256/80/80/22, 722.000 iterations,
W
hole face
(SFW -
o
ld
pre
train dataset
)
LIAE-UD
- res: 224, dims: 256/64/64/16, 1.000.000 iterations, Whole face
(SFW -
o
ld
pre
train dataset
)
@
DF-U
- res: 224, dims: 256/64/64/22, 412.000 iterations, Whole face
(SFW - FFHQ pretrain dataset)
DF-UD
- res: 384, dims: 256/64/64/22, 1.200.000 iterations,
Whole face
(SFW -
FFHQ p
retrain dataset
)
DF-UD
- res: 320, dims: 320/96/96/32, 560.000 iterations,
Whole face
(SFW - FFHQ pretrain dataset)
@
DF-UD
- res: 224, dims: 320/96/96/22, 428.817 iterations, Whole face
(SFW - FFHQ pretrain dataset)
DF-UD
- res: 256, dims: 292/92/92/16, 750.000 iterations, Whole face
(SFW - FFHQ pretrain dataset)
@
DF
- res 288, dims: 384/92/72/22, 1.000.000 iterations, Whole face
(SFW - old pretrain dataset)
@
DF
- res: 320, dims: 384/92/72/22, 419.891 iterations, Whole face
(SFW - old pretrain dataset)
@
DF-UD
- res: 320, dims: 256/64/64/22, 303.693 iterations, Whole Face (SFW - old pretrain dataset)
DF-UD
- res: 320, dims: 256/64/64/22, 1.000.000 iterations, Whole Face
(SFW - old pretrain dataset 250k + FFHQ set 850k)
@
DF-UD
- res: 256,
dims: 320/96/96/22, 300.000 iterations, Whole face (SFW - old pretrain dataset)
DF-UD - res: 192, dims: 256/64/64/22, 609.000 iterations, Whole face (SFW - FFHQ Female Only)
DF-UD
- res: 192, dims: 288/80/64/22, 546.000 iterations, Whole face
(SFW - FFHQ Female Only)
DF-UD - res: 224, dims: 256/64/64/22, 510.000 iterations, Whole face (SFW - FFHQ Female Only)
DF-UD - res: 256, dims: 256/64/64/22, 1.000.000 iterations, Whole face (SFW - FFHQ pretrain dataset)
@
DF-UD
- res: 512, dims: 256/64/80/22, 500.000 iterations, Whole face
(SFW - FFHQ pretrain dataset + XSeg)
Head:
NOTE: If you're pretraining a HEAD model consider using your custom pretrain set as the included FFHQ dataset is WF (whole face) only so it doesn't cover the whole head area needed to properly train HEAD face_type models. Old pretrain dataset is also not suitable as it's only FF (full face).
@
DF-UD
- res: 320, dims: 256/64/80/32, 1.000.000 iterations, Head
(SFW -
o
ld
pre
train dataset
)
@
DF-UD
- res: 448, dims: 256/64/64/16, 1.000.817 iterations, Head
(SFW - FFHQ
pre
train dataset
)
DF-UD - res: 224, dims: 256/64/64/22, 660.000 iterations, Head (SFW - FFHQ Female Only)
NSFW(用了有颜色数据训练的模型)models:
These are the models that are either pretrained using an
alternative dataset (such as one that focuses on one race, gender
or
contains NSFW samples from porn)
as well as
models that were prepared by training model on dataset of random faces with pretrainining disabled (regular training).
__________________________________________________
Half Face:
@
DF
- res: 192, dims: 256/64/64/22, 484.000 iterations, Half face
(
NSFW dataset - regular training -
women only)
Mid-Half Face:
@
DF
- res: 192, dims: 256/64/64/22, 241.307 iterations, Mid-half face
(
NSFW datas
et - regular training
-
women only)
Full Face:
@
DF
- res: 128, dims: 256/64/64/16, 129.388 iterations, Full face
(NSFW dataset - regular training - women only)
Fappenberg model reuploads:
DF
- res: 192, dims: 256/64/64/22, 300.000 iterations, Full face
(Pretrain 50k
+ NSFW dataset 250k regular training - women only)
LIAE-U
-
res: 192, dims:
256/64/64/22 dims, 150.
000 iterations, Full face
(100K pretrain
+ 50K custom female pretrain)
@
DF-U
- res: 192, dims: 256/64/64/22, 899.021 iterations, Full face
(Pretrain with
custo
m SFW dataset -
women only)
@
DF
- res: 160, dims: 256/64/64/22, 235.000 iterations, Full face
(custom SFW dataset - women/men)
LIAE
- res: 160, dims: 264/72/72/24, 244.000 iterations, Full face
(
custo
m SFW dataset -
women/men)
@
DF-U
- res: 192, dims: 256/64/64/22, 200.000 iterations, Full face
(
NSFW dataset- Pretrain with women only)
DF-U
-
res:
256, dims: 256/64/64/22,
285.000 iterations, Full face
(NSFW dataset - Pretrain with women only)
DF-UD
-res: 256, dims: 352/88/88/28, 250.000 iterations, Full face (NSFW dataset - Pretrain with women only)
DF-UD
-res: 256
, dims: 352/88/88/28, 698.000 iterations, Full face
(NSFW dataset - regular training - women only)
DF-UD -res: 320, dims: 360/90/90/22, 260.000 iterations, Full face (NSFW dataset - Pretrain with women only)
@
LIAE-U
- res: 192, dims: 256/64/64/22, 400.508 iterations, Full face
(NSFW dataset - Pretrain with women only)
Whole Face:
@
DF-U
- res: 288, dims: 256/64/64/22, 300.000 iterations, Whole face
(NSFW dataset - Pretrain with women only)
Head:
NOTE: If you're pretraining a HEAD model consider using your custom pretrain set as the included FFHQ dataset is WF (whole face) only so it doesn't cover the whole head area needed to properly train HEAD face_type models. Old pretrain dataset is also not suitable as it's only FF (full face).