def __init__(self, phase, base, extras, head, num_classes):
super(SSD, self).__init__()
self.phase = phase
self.num_classes = num_classes
# TODO: implement __call__ in PriorBox
self.priorbox = PriorBox(v2)
self.priors = self.priorbox.forward()
self.size = 300
# Layer learns to scale the l2 normalized features from conv4_3
self.L2Norm = L2Norm(512, 20)
# SSD network
self.vgg = nn.ModuleList(base)
self.extras = nn.ModuleList(extras)
self.loc = nn.ModuleList(head[0])
self.conf = nn.ModuleList(head[1])
if phase == 'test':
self.softmax = nn.Softmax()
self.detect = Detect(21, 0, 200, 0.01, 0.25, 400)
def __init__(self, phase, version, sz=300, num_classes=21):
super(SSD, self).__init__()
self.phase = phase
self.size = sz
self.num_classes = num_classes
param = num_classes * 3
self.base = build_base(cfg[str(sz)], 3)
self.version = v1
self.box_layer = PriorBox(self.version)
self.priors = Variable(self.box_layer.forward())
# TODO: Build the rest of the sequentials in a for loop.
self.features2 = nn.Sequential(
nn.MaxPool2d(kernel_size=2, stride=2),
nn.Conv2d(512, 512, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(512, 512, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(512, 512, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=1, padding=1),
nn.Conv2d(512, 1024, kernel_size=3, padding=6, dilation=6),
nn.ReLU(inplace=True),
nn.Conv2d(1024, 1024, kernel_size=1),
nn.ReLU(inplace=True),
)
self.features3 = nn.Sequential(
nn.Conv2d(1024, 256, kernel_size=1),
nn.ReLU(inplace=True),
nn.Conv2d(256, 512, kernel_size=3, stride=2, padding=1),
nn.ReLU(inplace=True),
)
self.features4 = nn.Sequential(
nn.Conv2d(512, 128, kernel_size=1),
nn.ReLU(inplace=True),
nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1),
nn.ReLU(inplace=True),
)
self.features5 = nn.Sequential(
nn.Conv2d(256, 128, kernel_size=1),
nn.ReLU(inplace=True),
nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1),
nn.ReLU(inplace=True),
)
self.pool6 = nn.Sequential(nn.AvgPool2d(kernel_size=3, stride=1), )
# Multibox layers (conv layers to learn features from different scales)
self.L2Norm = L2Norm(512, 20)
self.l4_3 = nn.Conv2d(512, 12, kernel_size=3, padding=1)
self.c4_3 = nn.Conv2d(512, param, kernel_size=3, padding=1)
self.lfc7 = nn.Conv2d(1024, 24, kernel_size=3, padding=1)
self.cfc7 = nn.Conv2d(1024, param * 2, kernel_size=3, padding=1)
self.l6_2 = nn.Conv2d(512, 24, kernel_size=3, padding=1)
self.c6_2 = nn.Conv2d(512, param * 2, kernel_size=3, padding=1)
self.l7_2 = nn.Conv2d(256, 24, kernel_size=3, padding=1)
self.c7_2 = nn.Conv2d(256, param * 2, kernel_size=3, padding=1)
self.l8_2 = nn.Conv2d(256, 24, kernel_size=3, padding=1)
self.c8_2 = nn.Conv2d(256, param * 2, kernel_size=3, padding=1)
self.lp6 = nn.Conv2d(256, 24, kernel_size=3, padding=1)
self.cp6 = nn.Conv2d(256, param * 2, kernel_size=3, padding=1)
self.softmax = nn.Softmax()
self.detect = Detect(21, 0, 200, 0.01, 0.45, 400)
class SSD(nn.Module):
"""Single Shot Multibox Architecture
The network is composed of a base VGG network followed by the
added multibox conv layers. Each multibox layer branches into
1) conv2d for class conf scores
2) conv2d for localization predictions
3) associated priorbox layer to produce default bounding
boxes specific to the layer's feature map size.
See: https://arxiv.org/pdf/1512.02325.pdf for more details.
Args:
features1: (nn.Sequential) VGG layers for input
size of either 300 or 512. Default: 300
phase: (string) Can be "test" or "train"
size: (int) the SSD version for the input size. Can be 300 or 500.
Defaul: 300
num_classes: (int) the number of classes to score. Default: 21.
"""
def __init__(self, phase, version, sz=300, num_classes=21):
super(SSD, self).__init__()
self.phase = phase
self.size = sz
self.num_classes = num_classes
param = num_classes * 3
self.base = build_base(cfg[str(sz)], 3)
self.version = v1
self.box_layer = PriorBox(self.version)
self.priors = Variable(self.box_layer.forward())
# TODO: Build the rest of the sequentials in a for loop.
self.features2 = nn.Sequential(
nn.MaxPool2d(kernel_size=2, stride=2),
nn.Conv2d(512, 512, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(512, 512, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(512, 512, kernel_size=3, padding=1),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=1, padding=1),
nn.Conv2d(512, 1024, kernel_size=3, padding=6, dilation=6),
nn.ReLU(inplace=True),
nn.Conv2d(1024, 1024, kernel_size=1),
nn.ReLU(inplace=True),
)
self.features3 = nn.Sequential(
nn.Conv2d(1024, 256, kernel_size=1),
nn.ReLU(inplace=True),
nn.Conv2d(256, 512, kernel_size=3, stride=2, padding=1),
nn.ReLU(inplace=True),
)
self.features4 = nn.Sequential(
nn.Conv2d(512, 128, kernel_size=1),
nn.ReLU(inplace=True),
nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1),
nn.ReLU(inplace=True),
)
self.features5 = nn.Sequential(
nn.Conv2d(256, 128, kernel_size=1),
nn.ReLU(inplace=True),
nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1),
nn.ReLU(inplace=True),
)
self.pool6 = nn.Sequential(nn.AvgPool2d(kernel_size=3, stride=1), )
# Multibox layers (conv layers to learn features from different scales)
self.L2Norm = L2Norm(512, 20)
self.l4_3 = nn.Conv2d(512, 12, kernel_size=3, padding=1)
self.c4_3 = nn.Conv2d(512, param, kernel_size=3, padding=1)
self.lfc7 = nn.Conv2d(1024, 24, kernel_size=3, padding=1)
self.cfc7 = nn.Conv2d(1024, param * 2, kernel_size=3, padding=1)
self.l6_2 = nn.Conv2d(512, 24, kernel_size=3, padding=1)
self.c6_2 = nn.Conv2d(512, param * 2, kernel_size=3, padding=1)
self.l7_2 = nn.Conv2d(256, 24, kernel_size=3, padding=1)
self.c7_2 = nn.Conv2d(256, param * 2, kernel_size=3, padding=1)
self.l8_2 = nn.Conv2d(256, 24, kernel_size=3, padding=1)
self.c8_2 = nn.Conv2d(256, param * 2, kernel_size=3, padding=1)
self.lp6 = nn.Conv2d(256, 24, kernel_size=3, padding=1)
self.cp6 = nn.Conv2d(256, param * 2, kernel_size=3, padding=1)
self.softmax = nn.Softmax()
self.detect = Detect(21, 0, 200, 0.01, 0.45, 400)
def forward(self, x):
"""Applies network layers and ops on input image(s) x.
Args:
x: input image or batch of images. Shape: [batch,3*batch,300,300].
Return:
Depending on phase:
test:
Variable(tensor) of output class label predictions,
confidence score, and corresponding location predictions for
each object detected. Shape: [batch,topk,7]
train:
list of concat outputs from:
1: confidence layers, Shape: [batch*num_priors,num_classes]
2: localization layers, Shape: [batch,num_priors*4]
3: priorbox layers, Shape: [2,num_priors*4]
"""
x = self.base(x)
y = self.L2Norm(x)
b1 = [
self.l4_3(y).permute(0, 2, 3, 1),
self.c4_3(y).permute(0, 2, 3, 1)
]
b1 = [o.view(o.contiguous().size(0), -1) for o in b1]
x = self.features2(x)
b2 = [
self.lfc7(x).permute(0, 2, 3, 1),
self.cfc7(x).permute(0, 2, 3, 1)
]
b2 = [o.view(o.contiguous().size(0), -1) for o in b2]
x = self.features3(x)
b3 = [
self.l6_2(x).permute(0, 2, 3, 1),
self.c6_2(x).permute(0, 2, 3, 1)
]
b3 = [o.view(o.contiguous().size(0), -1) for o in b3]
x = self.features4(x)
b4 = [
self.l7_2(x).permute(0, 2, 3, 1),
self.c7_2(x).permute(0, 2, 3, 1)
]
b4 = [o.view(o.contiguous().size(0), -1) for o in b4]
x = self.features5(x)
b5 = [
self.l8_2(x).permute(0, 2, 3, 1),
self.c8_2(x).permute(0, 2, 3, 1)
]
b5 = [o.view(o.contiguous().size(0), -1) for o in b5]
x = self.pool6(x)
b6 = [self.lp6(x).permute(0, 2, 3, 1), self.cp6(x).permute(0, 2, 3, 1)]
b6 = [o.view(o.contiguous().size(0), -1) for o in b6]
loc = torch.cat((b1[0], b2[0], b3[0], b4[0], b5[0], b6[0]), 1)
conf = torch.cat((b1[1], b2[1], b3[1], b4[1], b5[1], b6[1]), 1)
if self.phase == "test":
output = self.detect(
loc.view(loc.size(0), -1, 4), # loc preds
self.softmax(conf.view(-1, self.num_classes)), # conf preds
self.priors # default boxes
)
else:
print(self.priors.size())
conf = conf.view(conf.size(0), -1, self.num_classes)
loc = loc.view(loc.size(0), -1, 4)
output = (loc, conf, self.priors)
return output
def load_weights(self, base_file):
other, ext = os.path.splitext(base_file)
if ext == '.pkl' or '.pth':
print('Loading weights into state dict...')
self.load_state_dict(torch.load(base_file))
print('Finished!')
else:
print('Error: Sorry Only .pth and .pkl files currently supported!')
class SSD(nn.Module):
"""Single Shot Multibox Architecture
The network is composed of a base VGG network followed by the
added multibox conv layers. Each multibox layer branches into
1) conv2d for class conf scores
2) conv2d for localization predictions
3) associated priorbox layer to produce default bounding
boxes specific to the layer's feature map size.
See: https://arxiv.org/pdf/1512.02325.pdf for more details.
Args:
phase: (string) Can be "test" or "train"
base: VGG16 layers for input, size of either 300 or 500
extras: extra layers that feed to multibox loc and conf layers
head: "multibox head" consists of loc and conf conv layers
"""
def __init__(self, phase, base, extras, head, num_classes):
super(SSD, self).__init__()
self.phase = phase
self.num_classes = num_classes
# TODO: implement __call__ in PriorBox
self.priorbox = PriorBox(v2)
self.priors = self.priorbox.forward()
self.size = 300
# Layer learns to scale the l2 normalized features from conv4_3
self.L2Norm = L2Norm(512, 20)
# SSD network
self.vgg = nn.ModuleList(base)
self.extras = nn.ModuleList(extras)
self.loc = nn.ModuleList(head[0])
self.conf = nn.ModuleList(head[1])
if phase == 'test':
self.softmax = nn.Softmax()
self.detect = Detect(21, 0, 200, 0.01, 0.25, 400)
def forward(self, x):
"""Applies network layers and ops on input image(s) x.
Args:
x: input image or batch of images. Shape: [batch,3*batch,300,300].
Return:
Depending on phase:
test:
Variable(tensor) of output class label predictions,
confidence score, and corresponding location predictions for
each object detected. Shape: [batch,topk,7]
train:
list of concat outputs from:
1: confidence layers, Shape: [batch*num_priors,num_classes]
2: localization layers, Shape: [batch,num_priors*4]
3: priorbox layers, Shape: [2,num_priors*4]
"""
sources = list()
loc = list()
conf = list()
# apply vgg up to conv4_3 relu
for k in range(23):
x = self.vgg[k](x)
s = self.L2Norm(x)
sources.append(s)
# apply vgg up to fc7
for k in range(23, len(self.vgg)):
x = self.vgg[k](x)
sources.append(x)
# apply extra layers and cache source layer outputs
for k, v in enumerate(self.extras):
x = F.relu(v(x), inplace=True)
if k % 2 == 1:
sources.append(x)
# apply multibox head to source layers
for (x, l, c) in zip(sources, self.loc, self.conf):
loc.append(l(x).permute(0, 2, 3, 1).contiguous())
conf.append(c(x).permute(0, 2, 3, 1).contiguous())
loc = torch.cat([o.view(o.size(0), -1) for o in loc], 1)
conf = torch.cat([o.view(o.size(0), -1) for o in conf], 1)
if self.phase == "test":
output = self.detect(
loc.view(loc.size(0), -1, 4), # loc preds
self.softmax(conf.view(-1, self.num_classes)), # conf preds
Variable(self.priors, volatile=True) # default boxes
)
else:
conf = conf.view(conf.size(0), -1, self.num_classes)
loc = loc.view(loc.size(0), -1, 4)
output = (loc, conf, self.priors)
return output
def load_weights(self, base_file):
other, ext = os.path.splitext(base_file)
if ext == '.pkl' or '.pth':
print('Loading weights into state dict...')
self.load_state_dict(torch.load(base_file))
print('Finished!')
else:
print('Sorry only .pth and .pkl files supported.')