def r2plus1d_152(pretrained=False, progress=False, **kwargs):
return _video_resnet('r2plus1d_152',
False,
False,
block=Bottleneck,
conv_makers=[Conv2Plus1D] * 4,
layers=[3, 8, 36, 3],
stem=R2Plus1dStem,
**kwargs)
def r2plus1d_34(pretrained=False, progress=False, **kwargs):
return _video_resnet('r2plus1d_34',
False,
False,
block=BasicBlock,
conv_makers=[Conv2Plus1D] * 4,
layers=[3, 4, 6, 3],
stem=R2Plus1dStem,
**kwargs)
def r2plus1d_18(pretrained=False, progress=False, **kwargs):
"""Constructs a ResNet-18 model.
"""
return _video_resnet('r2plus1d_18',
False,
False,
block=BasicBlock,
conv_makers=[Conv2Plus1D] * 4,
layers=[2, 2, 2, 2],
stem=R2Plus1dStem,
**kwargs)
def ip_csn_50(pretrained=False, progress=False, **kwargs):
model = _video_resnet("ip_csn_50",
False,
False,
block=Bottleneck,
conv_makers=[IPConv3DDepthwise] * 4,
layers=[3, 8, 6, 3],
stem=BasicStem_Pool,
**kwargs)
for m in model.modules():
if isinstance(m, nn.BatchNorm3d):
m.eps = 1e-3
return model
def r2plus1d_152(pretrained=False, progress=False, **kwargs):
model = _video_resnet("r2plus1d_152",
False,
False,
block=Bottleneck,
conv_makers=[Conv2Plus1D] * 4,
layers=[3, 8, 36, 3],
stem=R2Plus1dStem,
**kwargs)
for m in model.modules():
if isinstance(m, nn.BatchNorm3d):
m.eps = 1e-3
return model
def r2plus1d_18(pretrained=False, progress=False, **kwargs):
model = _video_resnet("r2plus1d_18",
False,
False,
block=BasicBlock,
conv_makers=[Conv2Plus1D] * 4,
layers=[2, 2, 2, 2],
stem=R2Plus1dStem,
**kwargs)
# We need exact Caffe2 momentum for BatchNorm scaling
for m in model.modules():
if isinstance(m, nn.BatchNorm3d):
m.eps = 1e-3
m.momentum = 0.9
return model
