def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.in_channels = 3
self.features[0] = BasicConv2d(self.in_channels,
32,
kernel_size=3,
stride=2,
padding=1)
self.features[1] = BasicConv2d(32,
32,
kernel_size=3,
stride=1,
padding=1)
self.chunks = [3, 5, 9, 15]
# correct paddings
for m in self.modules():
if isinstance(m, nn.Conv2d):
if m.kernel_size == (3, 3):
m.padding = (1, 1)
if isinstance(m, nn.MaxPool2d):
m.padding = (1, 1)
# remove linear layers
del self.last_linear
def __init__(self, config, anchors, num_cls):
nn.Module.__init__(self)
self.features = nn.Sequential(
BasicConv2d(3, 32, kernel_size=3, stride=2),
BasicConv2d(32, 32, kernel_size=3, stride=1),
BasicConv2d(32, 64, kernel_size=3, stride=1, padding=1),
Mixed_3a(),
Mixed_4a(),
Mixed_5a(),
Inception_A(),
Inception_A(),
Inception_A(),
Inception_A(),
Reduction_A(), # Mixed_6a
Inception_B(),
Inception_B(),
Inception_B(),
Inception_B(),
Inception_B(),
Inception_B(),
Inception_B(),
Reduction_B(), # Mixed_7a
Inception_C(),
Inception_C(),
Inception_C(),
nn.Conv2d(1536, model.output_channels(len(anchors), num_cls), 1),
)
gamma = config.getboolean('batch_norm', 'gamma')
beta = config.getboolean('batch_norm', 'beta')
for m in self.modules():
if isinstance(m, nn.Conv2d):
m.weight = nn.init.kaiming_normal(m.weight)
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
m.weight.requires_grad = gamma
m.bias.requires_grad = beta
if config.getboolean('model', 'pretrained'):
settings = pretrained_settings['inceptionv4'][config.get('inception4', 'pretrained')]
logging.info('use pretrained model: ' + str(settings))
state_dict = self.state_dict()
for key, value in torch.utils.model_zoo.load_url(settings['url']).items():
if key in state_dict:
state_dict[key] = value
self.load_state_dict(state_dict)
def __init__(self, config_channels, prefix):
nn.Module.__init__(self)
channels = config_channels.channels
self.maxpool = nn.MaxPool2d(3, stride=2)
self.conv = BasicConv2d(config_channels.channels,
config_channels(96, '%s.conv.conv.weight' %
prefix),
kernel_size=3,
stride=2)
config_channels.channels = channels + self.conv.conv.weight.size(0)
def __init__(self, config_channels, prefix):
nn.Module.__init__(self)
channels = config_channels.channels
self.branch0 = BasicConv2d(config_channels.channels,
config_channels(
384, '%s.branch0.conv.weight' % prefix),
kernel_size=3,
stride=2)
# branch1
config_channels.channels = channels
branch = []
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
192, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=1,
stride=1))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
224, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=3,
stride=1,
padding=1))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
256, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=3,
stride=2))
self.branch1 = nn.Sequential(*branch)
self.branch2 = nn.MaxPool2d(3, stride=2)
# output
config_channels.channels = self.branch0.conv.weight.size(
0) + self.branch1[-1].conv.weight.size(0) + channels
def __init__(self, config_channels, prefix):
nn.Module.__init__(self)
# branch0
channels = config_channels.channels
branch = []
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
64, '%s.branch0.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=1,
stride=1))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
96, '%s.branch0.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=3,
stride=1))
self.branch0 = nn.Sequential(*branch)
# branch1
config_channels.channels = channels
branch = []
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
64, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=1,
stride=1))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
64, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(1, 7),
stride=1,
padding=(0, 3)))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
64, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(7, 1),
stride=1,
padding=(3, 0)))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
96, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(3, 3),
stride=1))
self.branch1 = nn.Sequential(*branch)
# output
config_channels.channels = self.branch0[-1].conv.weight.size(
0) + self.branch1[-1].conv.weight.size(0)
def __init__(self, config_channels, anchors, num_cls):
nn.Module.__init__(self)
features = []
features.append(
BasicConv2d(config_channels.channels,
config_channels(
32, 'features.%d.conv.weight' % len(features)),
kernel_size=3,
stride=2))
features.append(
BasicConv2d(config_channels.channels,
config_channels(
32, 'features.%d.conv.weight' % len(features)),
kernel_size=3,
stride=1))
features.append(
BasicConv2d(config_channels.channels,
config_channels(
64, 'features.%d.conv.weight' % len(features)),
kernel_size=3,
stride=1,
padding=1))
features.append(
Mixed_3a(config_channels, 'features.%d' % len(features)))
features.append(
Mixed_4a(config_channels, 'features.%d' % len(features)))
features.append(
Mixed_5a(config_channels, 'features.%d' % len(features)))
features.append(
Inception_A(config_channels, 'features.%d' % len(features)))
features.append(
Inception_A(config_channels, 'features.%d' % len(features)))
features.append(
Inception_A(config_channels, 'features.%d' % len(features)))
features.append(
Inception_A(config_channels, 'features.%d' % len(features)))
features.append(
Reduction_A(config_channels,
'features.%d' % len(features))) # Mixed_6a
features.append(
Inception_B(config_channels, 'features.%d' % len(features)))
features.append(
Inception_B(config_channels, 'features.%d' % len(features)))
features.append(
Inception_B(config_channels, 'features.%d' % len(features)))
features.append(
Inception_B(config_channels, 'features.%d' % len(features)))
features.append(
Inception_B(config_channels, 'features.%d' % len(features)))
features.append(
Inception_B(config_channels, 'features.%d' % len(features)))
features.append(
Inception_B(config_channels, 'features.%d' % len(features)))
features.append(
Reduction_B(config_channels,
'features.%d' % len(features))) # Mixed_7a
features.append(
Inception_C(config_channels, 'features.%d' % len(features)))
features.append(
Inception_C(config_channels, 'features.%d' % len(features)))
features.append(
Inception_C(config_channels, 'features.%d' % len(features)))
features.append(
nn.Conv2d(config_channels.channels,
model.output_channels(len(anchors), num_cls), 1))
self.features = nn.Sequential(*features)
try:
gamma = config_channels.config.getboolean('batch_norm', 'gamma')
except (configparser.NoSectionError, configparser.NoOptionError):
gamma = True
try:
beta = config_channels.config.getboolean('batch_norm', 'beta')
except (configparser.NoSectionError, configparser.NoOptionError):
beta = True
for m in self.modules():
if isinstance(m, nn.Conv2d):
m.weight = nn.init.kaiming_normal(m.weight)
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
m.weight.requires_grad = gamma
m.bias.requires_grad = beta
try:
if config_channels.config.getboolean('model', 'pretrained'):
settings = pretrained_settings['inceptionv4'][
config_channels.config.get('inception4', 'pretrained')]
logging.info('use pretrained model: ' + str(settings))
state_dict = self.state_dict()
for key, value in torch.utils.model_zoo.load_url(
settings['url']).items():
if key in state_dict:
state_dict[key] = value
self.load_state_dict(state_dict)
except (configparser.NoSectionError, configparser.NoOptionError):
pass
def __init__(self, config_channels, prefix):
nn.Module.__init__(self)
channels = config_channels.channels
self.branch0 = BasicConv2d(config_channels.channels,
config_channels(
256, '%s.branch0.conv.weight' % prefix),
kernel_size=1,
stride=1)
# branch1
config_channels.channels = channels
self.branch1_0 = BasicConv2d(config_channels.channels,
config_channels(
384,
'%s.branch1_0.conv.weight' % prefix),
kernel_size=1,
stride=1)
_channels = config_channels.channels
self.branch1_1a = BasicConv2d(
_channels,
config_channels(256, '%s.branch1_1a.conv.weight' % prefix),
kernel_size=(1, 3),
stride=1,
padding=(0, 1))
self.branch1_1b = BasicConv2d(
_channels,
config_channels(256, '%s.branch1_1b.conv.weight' % prefix),
kernel_size=(3, 1),
stride=1,
padding=(1, 0))
# branch2
config_channels.channels = channels
self.branch2_0 = BasicConv2d(config_channels.channels,
config_channels(
384,
'%s.branch2_0.conv.weight' % prefix),
kernel_size=1,
stride=1)
self.branch2_1 = BasicConv2d(config_channels.channels,
config_channels(
448,
'%s.branch2_1.conv.weight' % prefix),
kernel_size=(3, 1),
stride=1,
padding=(1, 0))
self.branch2_2 = BasicConv2d(config_channels.channels,
config_channels(
512,
'%s.branch2_2.conv.weight' % prefix),
kernel_size=(1, 3),
stride=1,
padding=(0, 1))
_channels = config_channels.channels
self.branch2_3a = BasicConv2d(
_channels,
config_channels(256, '%s.branch2_3a.conv.weight' % prefix),
kernel_size=(1, 3),
stride=1,
padding=(0, 1))
self.branch2_3b = BasicConv2d(
_channels,
config_channels(256, '%s.branch2_3b.conv.weight' % prefix),
kernel_size=(3, 1),
stride=1,
padding=(1, 0))
# branch3
config_channels.channels = channels
branch = []
branch.append(
nn.AvgPool2d(3, stride=1, padding=1, count_include_pad=False))
branch.append(
BasicConv2d(config_channels.channels, 256, kernel_size=1,
stride=1))
self.branch3 = nn.Sequential(*branch)
# output
config_channels.channels = self.branch0.conv.weight.size(
0) + self.branch1_1a.conv.weight.size(
0) + self.branch1_1b.conv.weight.size(
0) + self.branch2_3a.conv.weight.size(
0) + self.branch2_3b.conv.weight.size(
0) + self.branch3[-1].conv.weight.size(0)
def __init__(self, config_channels, prefix):
nn.Module.__init__(self)
channels = config_channels.channels
self.branch0 = BasicConv2d(config_channels.channels,
config_channels(
384, '%s.branch0.conv.weight' % prefix),
kernel_size=1,
stride=1)
# branch1
config_channels.channels = channels
branch = []
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
192, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=1,
stride=1))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
224, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(1, 7),
stride=1,
padding=(0, 3)))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
256, '%s.branch1.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(7, 1),
stride=1,
padding=(3, 0)))
self.branch1 = nn.Sequential(*branch)
# branch2
config_channels.channels = channels
branch = []
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
192, '%s.branch2.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=1,
stride=1))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
192, '%s.branch2.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(7, 1),
stride=1,
padding=(3, 0)))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
224, '%s.branch2.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(1, 7),
stride=1,
padding=(0, 3)))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
224, '%s.branch2.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(7, 1),
stride=1,
padding=(3, 0)))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
256, '%s.branch2.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=(1, 7),
stride=1,
padding=(0, 3)))
self.branch2 = nn.Sequential(*branch)
# branch3
config_channels.channels = channels
branch = []
branch.append(
nn.AvgPool2d(3, stride=1, padding=1, count_include_pad=False))
branch.append(
BasicConv2d(config_channels.channels,
config_channels(
128, '%s.branch3.%d.conv.weight' %
(prefix, len(branch))),
kernel_size=1,
stride=1))
self.branch3 = nn.Sequential(*branch)
# output
config_channels.channels = self.branch0.conv.weight.size(
0) + self.branch1[-1].conv.weight.size(0) + self.branch2[
-1].conv.weight.size(0) + self.branch3[-1].conv.weight.size(0)