def hardnet_85(pretrained=False, return_transforms=False, **kwargs):
model = HarDNet(arch=85, **kwargs)
if pretrained:
model = load_model(model, urls["hardnet_85"])
if return_transforms:
return model, transforms
else:
return model
def rexnet_3_0(pretrained=False, return_transforms=False, **kwargs):
model = ReXNet(width_mult=3.0, **kwargs)
if pretrained:
model = load_model(model, urls["rexnet_3_0"])
if return_transforms:
return model, transforms
else:
return model
def hardnet_68_ds(pretrained=False, return_transforms=False, **kwargs):
model = HarDNet(arch=68, depth_wise=True, **kwargs)
if pretrained:
model = load_model(model, urls["hardnet_68_ds"])
if return_transforms:
return model, transforms
else:
return model
def swin_s(pretrained=False, return_transforms=False, **kwargs):
model = SwinTransformer(depths=[2, 2, 18, 2], num_heads=[3, 6, 12, 24], **kwargs)
if pretrained:
model = load_model(model, urls["swin_s"])
if return_transforms:
return model, transforms_224
else:
return model
def swin_ti(pretrained=False, return_transforms=False, **kwargs):
model = SwinTransformer(**kwargs)
if pretrained:
model = load_model(model, urls["swin_ti"])
if return_transforms:
return model, transforms_224
else:
return model
def rednet_152(pretrained=False, return_transforms=False, **kwargs):
model = RedNet(BottleneckBlock, 152, **kwargs)
if pretrained:
model = load_model(model, urls["rednet_152"])
if return_transforms:
return model, transforms
else:
return model
def pvt_l(pretrained=False, return_transforms=False, **kwargs):
model = PyramidVisionTransformer(depths=[3, 8, 27, 3], **kwargs)
if pretrained:
model = load_model(model, urls["pvt_l"])
if return_transforms:
return model, transforms
else:
return model
def swin_b(pretrained=False, return_transforms=False, **kwargs):
model = SwinTransformer(
embed_dim=128, depths=[2, 2, 18, 2], num_heads=[4, 8, 16, 32], **kwargs
)
if pretrained:
model = load_model(model, urls["swin_b"])
if return_transforms:
return model, transforms_224
else:
return model
def dla_34(pretrained=False, return_transforms=False, **kwargs):
model = DLA(levels=(1, 1, 1, 2, 2, 1),
channels=(16, 32, 64, 128, 256, 512),
block=DlaBasic,
**kwargs)
if pretrained:
model = load_model(model, urls["dla_34"])
if return_transforms:
return model, transforms
else:
return model
def dla_60(pretrained=False, return_transforms=False, **kwargs):
model = DLA(levels=(1, 1, 1, 2, 3, 1),
channels=(16, 32, 128, 256, 512, 1024),
block=DlaBottleneck,
**kwargs)
if pretrained:
model = load_model(model, urls["dla_60"])
if return_transforms:
return model, transforms
else:
return model
def mixer_l(pretrained=False, return_transforms=False, **kwargs):
model = MlpMixer(hidden_dim=1024,
num_blocks=24,
tokens_mlp_dim=512,
channels_mlp_dim=4096,
**kwargs)
if pretrained:
model = load_model(model, urls["mixer_l"])
if return_transforms:
return model, transforms
else:
return model
def mixer_b(pretrained=False, return_transforms=False, **kwargs):
model = MlpMixer(hidden_dim=768,
num_blocks=12,
tokens_mlp_dim=384,
channels_mlp_dim=3072,
**kwargs)
if pretrained:
model = load_model(model, urls["mixer_b"])
if return_transforms:
return model, transforms
else:
return model
def repvgg_a2(pretrained=False, return_transforms=False, **kwargs):
model = RepVGG(
num_blocks=[2, 4, 14, 1],
width_multiplier=[1.5, 1.5, 1.5, 2.75],
override_groups_map=None,
**kwargs
)
if pretrained:
model = load_model(model, urls["repvgg_a2"])
if return_transforms:
return model, transforms
else:
return model
def t2t_vit_19(pretrained=False, return_transforms=False, **kwargs):
model = T2T_ViT(tokens_type="performer",
embed_dim=448,
depth=19,
num_heads=7,
mlp_ratio=3.0,
**kwargs)
if pretrained:
model = load_model(model, urls["t2t_vit_19"])
if return_transforms:
return model, transforms_224
else:
return model
def t2t_vit_t_24(pretrained=False, return_transforms=False, **kwargs):
model = T2T_ViT(tokens_type="transformer",
embed_dim=512,
depth=24,
num_heads=8,
mlp_ratio=3.0,
**kwargs)
if pretrained:
model = load_model(model, urls["t2t_vit_t_24"])
if return_transforms:
return model, transforms_224
else:
return model
def dla_60x_c(pretrained=False, return_transforms=False, **kwargs):
model = DLA(levels=(1, 1, 1, 2, 3, 1),
channels=(16, 32, 64, 64, 128, 256),
block=DlaBottleneck,
cardinality=32,
base_width=4,
**kwargs)
if pretrained:
model = load_model(model, urls["dla_60x_c"])
if return_transforms:
return model, transforms
else:
return model
def repvgg_b3g4(pretrained=False, return_transforms=False, **kwargs):
model = RepVGG(
num_blocks=[4, 6, 16, 1],
width_multiplier=[3, 3, 3, 5],
override_groups_map=g4_map,
**kwargs
)
if pretrained:
model = load_model(model, urls["repvgg_b3g4"])
if return_transforms:
return model, transforms
else:
return model
def dla_102x2(pretrained=False, return_transforms=False, **kwargs):
model = DLA(levels=(1, 1, 1, 3, 4, 1),
channels=(16, 32, 128, 256, 512, 1024),
block=DlaBottleneck,
cardinality=64,
base_width=4,
residual_root=True,
**kwargs)
if pretrained:
model = load_model(model, urls["dla_102x2"])
if return_transforms:
return model, transforms
else:
return model
def t2t_vit_14_384(pretrained=False, return_transforms=False, **kwargs):
model = T2T_ViT(img_size=384,
tokens_type="performer",
embed_dim=384,
depth=14,
num_heads=6,
mlp_ratio=3.0,
**kwargs)
if pretrained:
model = load_model(model, urls["t2t_vit_14_384"])
if return_transforms:
return model, transforms_384
else:
return model
def cait_m_48_448(pretrained=False, return_transforms=False, **kwargs):
model = CaiT(img_size=448,
embed_dim=768,
depth=48,
num_heads=16,
init_scale=1e-6,
**kwargs)
if pretrained:
model = load_model(model, urls["cait_m_48_448"])
if return_transforms:
return model, transforms_448
else:
return model
def cait_m_36_384(pretrained=False, return_transforms=False, **kwargs):
model = CaiT(img_size=384,
embed_dim=768,
depth=36,
num_heads=16,
init_scale=1e-6,
**kwargs)
if pretrained:
model = load_model(model, urls["cait_m_36_384"])
if return_transforms:
return model, transforms_384
else:
return model
def cait_xs_24_384(pretrained=False, return_transforms=False, **kwargs):
model = CaiT(img_size=384,
embed_dim=288,
depth=24,
num_heads=6,
init_scale=1e-5,
**kwargs)
if pretrained:
model = load_model(model, urls["cait_xs_24_384"])
if return_transforms:
return model, transforms_384
else:
return model
def cait_xxs_36(pretrained=False, return_transforms=False, **kwargs):
model = CaiT(img_size=224,
embed_dim=192,
depth=36,
num_heads=4,
init_scale=1e-5,
**kwargs)
if pretrained:
model = load_model(model, urls["cait_xxs_36"])
if return_transforms:
return model, transforms_224
else:
return model
def deit_b_distilled(pretrained=False, return_transforms=False, **kwargs):
model = DistilledVisionTransformer(patch_size=16,
embed_dim=768,
depth=12,
num_heads=12,
mlp_ratio=4,
qkv_bias=True,
epsilon=1e-6,
**kwargs)
if pretrained:
model = load_model(model, urls["deit_b_distilled"])
if return_transforms:
return model, transforms_224
else:
return model
def pit_b(pretrained=False, return_transforms=False, **kwargs):
model = PoolingTransformer(image_size=224,
patch_size=14,
stride=7,
base_dims=[64, 64, 64],
depth=[3, 6, 4],
heads=[4, 8, 16],
mlp_ratio=4,
**kwargs)
if pretrained:
model = load_model(model, urls["pit_b"])
if return_transforms:
return model, transforms
else:
return model
def pit_s(pretrained=False, return_transforms=False, **kwargs):
model = PoolingTransformer(image_size=224,
patch_size=16,
stride=8,
base_dims=[48, 48, 48],
depth=[2, 6, 4],
heads=[3, 6, 12],
mlp_ratio=4,
**kwargs)
if pretrained:
model = load_model(model, urls["pit_s"])
if return_transforms:
return model, transforms
else:
return model
def pit_ti_distilled(pretrained=False, return_transforms=False, **kwargs):
model = DistilledPoolingTransformer(image_size=224,
patch_size=16,
stride=8,
base_dims=[32, 32, 32],
depth=[2, 6, 4],
heads=[2, 4, 8],
mlp_ratio=4,
**kwargs)
if pretrained:
model = load_model(model, urls["pit_ti_distilled"])
if return_transforms:
return model, transforms
else:
return model
def swin_b_384(pretrained=False, return_transforms=False, **kwargs):
model = SwinTransformer(
img_size=384,
embed_dim=128,
depths=[2, 2, 18, 2],
num_heads=[4, 8, 16, 32],
window_size=12,
**kwargs,
)
if pretrained:
model = load_model(model, urls["swin_b_384"])
if return_transforms:
return model, transforms_384
else:
return model
def coat_m(pretrained=False, return_transforms=False, **kwargs):
model = CoaT(
patch_size=4,
embed_dims=[152, 216, 216, 216],
serial_depths=[2, 2, 2, 2],
parallel_depth=6,
num_heads=8,
mlp_ratios=[4, 4, 4, 4],
**kwargs,
)
if pretrained:
model = load_model(model, urls["coat_m"])
if return_transforms:
return model, transforms
else:
return model
def coat_lite_s(pretrained=False, return_transforms=False, **kwargs):
model = CoaT(
patch_size=4,
embed_dims=[64, 128, 320, 512],
serial_depths=[3, 4, 6, 3],
parallel_depth=0,
num_heads=8,
mlp_ratios=[8, 8, 4, 4],
**kwargs,
)
if pretrained:
model = load_model(model, urls["coat_lite_s"])
if return_transforms:
return model, transforms
else:
return model
