# MobileNet v2 derived from https://github.com/tonylins/pytorch-mobilenet-v2
import math
import torch.nn as nn
import thelper
[docs]def conv_bn(inp, oup, stride):
return nn.Sequential(
nn.Conv2d(inp, oup, 3, stride, 1, bias=False),
nn.BatchNorm2d(oup),
nn.ReLU6(inplace=True)
)
[docs]def conv_1x1_bn(inp, oup):
return nn.Sequential(
nn.Conv2d(inp, oup, 1, 1, 0, bias=False),
nn.BatchNorm2d(oup),
nn.ReLU6(inplace=True)
)
class InvertedResidual(nn.Module):
def __init__(self, inp, oup, stride, expand_ratio):
super().__init__()
self.stride = stride
if stride not in [1, 2]:
raise AssertionError("stride should be 1 or 2")
self.use_res_connect = self.stride == 1 and inp == oup
self.conv = nn.Sequential(
# pw
nn.Conv2d(inp, inp * expand_ratio, 1, 1, 0, bias=False),
nn.BatchNorm2d(inp * expand_ratio),
nn.ReLU6(inplace=True),
# dw
nn.Conv2d(inp * expand_ratio, inp * expand_ratio, 3, stride, 1, groups=inp * expand_ratio, bias=False),
nn.BatchNorm2d(inp * expand_ratio),
nn.ReLU6(inplace=True),
# pw-linear
nn.Conv2d(inp * expand_ratio, oup, 1, 1, 0, bias=False),
nn.BatchNorm2d(oup),
)
def forward(self, x):
if self.use_res_connect:
return x + self.conv(x)
else:
return self.conv(x)
class MobileNetV2(thelper.nn.Module):
def __init__(self, task, input_size=224, width_mult=1.):
# note: must always forward args to base class to keep backup
super().__init__(task, input_size=input_size, width_mult=width_mult)
# setting of inverted residual blocks
self.interverted_residual_setting = [
# t, c, n, s
[1, 16, 1, 1],
[6, 24, 2, 2],
[6, 32, 3, 2],
[6, 64, 4, 2],
[6, 96, 3, 1],
[6, 160, 3, 2],
[6, 320, 1, 1],
]
# building first layer
if input_size % 32 != 0:
raise AssertionError("input size should be multiple of 32 px")
input_channel = int(32 * width_mult)
self.last_channel = int(1280 * width_mult) if width_mult > 1.0 else 1280
self.features = [conv_bn(3, input_channel, 2)]
# building inverted residual blocks
for t, c, n, s in self.interverted_residual_setting:
output_channel = int(c * width_mult)
for i in range(n):
if i == 0:
self.features.append(InvertedResidual(input_channel, output_channel, s, t))
else:
self.features.append(InvertedResidual(input_channel, output_channel, 1, t))
input_channel = output_channel
# building last several layers
self.features.append(conv_1x1_bn(input_channel, self.last_channel))
self.features.append(nn.AvgPool2d(input_size // 32))
# make it nn.Sequential
self.features = nn.Sequential(*self.features)
# building classifier
self.classif_features = 1000 # default count, will be updated if needed
self.classifier = nn.Sequential(
nn.Dropout(),
nn.Linear(self.last_channel, self.classif_features),
)
self._initialize_weights()
self.set_task(task)
def forward(self, x):
x = self.features(x)
x = x.view(-1, self.last_channel)
x = self.classifier(x)
return x
def set_task(self, task):
assert isinstance(task, thelper.tasks.Classification), "missing impl for non-classif task type"
num_classes = len(task.class_names)
if num_classes != self.classif_features:
self.classifier = nn.Sequential(
nn.Dropout(),
nn.Linear(self.last_channel, num_classes),
)
self.classif_features = num_classes
self.task = task
def _initialize_weights(self):
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
if m.bias is not None:
m.bias.data.zero_()
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
elif isinstance(m, nn.Linear):
n = m.weight.size(1)
m.weight.data.normal_(0, 0.01)
m.bias.data.zero_()
