def create_network(self, blocks):
hyperparams = blocks[0]
output_filters = [int(hyperparams['channels'])]
module_list = nn.ModuleList()
routs = [] # list of layers which rout to deeper layers
ind = -2
filters = -1
for mdef in blocks:
ind += 1
modules = nn.Sequential()
if mdef['type'] in ['net', 'learnet']:
continue
if mdef['type'] == 'convolutional':
bn = int(mdef['batch_normalize'])
filters = int(mdef['filters'])
size = int(mdef['size'])
stride = int(mdef['stride']) if 'stride' in mdef else (int(mdef['stride_y']), int(mdef['stride_x']))
pad = (size - 1) // 2 if int(mdef['pad']) else 0
dynamic = True if 'dynamic' in mdef and int(mdef['dynamic']) == 1 else False
if dynamic:
partial = int(mdef['partial']) if 'partial' in mdef else None
Conv2d = dynamic_conv2d(is_first=True, partial=partial)
else:
Conv2d = nn.Conv2d
modules.add_module('Conv2d', Conv2d(in_channels=output_filters[-1],
out_channels=filters,
kernel_size=size,
stride=stride,
padding=pad,
groups=int(mdef['groups']) if 'groups' in mdef else 1,
bias=not bn))
if bn:
modules.add_module('BatchNorm2d', nn.BatchNorm2d(filters, momentum=0.1))
if mdef['activation'] == 'leaky':
# TODO: activation study https://github.com/ultralytics/yolov3/issues/441
modules.add_module('activation', nn.LeakyReLU(0.1, inplace=True))
# modules.add_module('activation', nn.PReLU(num_parameters=1, init=0.10))
elif mdef['activation'] == 'swish':
modules.add_module('activation', Swish())
elif mdef['type'] == 'maxpool':
size = int(mdef['size'])
stride = int(mdef['stride'])
maxpool = nn.MaxPool2d(kernel_size=size, stride=stride, padding=int((size - 1) // 2))
if size == 2 and stride == 1: # yolov3-tiny
modules.add_module('ZeroPad2d', nn.ZeroPad2d((0, 1, 0, 1)))
modules.add_module('MaxPool2d', maxpool)
else:
modules = maxpool
elif mdef['type'] == 'upsample':
modules = nn.Upsample(scale_factor=int(mdef['stride']), mode='nearest')
elif mdef['type'] == 'route': # nn.Sequential() placeholder for 'route' layer
layers = [int(x) for x in mdef['layers'].split(',')]
filters = sum([output_filters[i + 1 if i > 0 else i] for i in layers])
routs.extend([l if l > 0 else l + ind for l in layers])
modules = EmptyModule()
# if mdef[i+1]['type'] == 'reorg3d':
# modules = nn.Upsample(scale_factor=1/float(mdef[i+1]['stride']), mode='nearest') # reorg3d
elif mdef['type'] == 'shortcut': # nn.Sequential() placeholder for 'shortcut' layer
filters = output_filters[int(mdef['from'])]
layer = int(mdef['from'])
routs.extend([ind + layer if layer < 0 else layer])
modules = EmptyModule()
elif mdef['type'] == 'region':
loss = RegionLossV2()
anchors = mdef['anchors'].split(',')
loss.anchors = [float(i) for i in anchors]
loss.num_classes = int(mdef['classes'])
loss.num_anchors = int(mdef['num'])
loss.object_scale = float(mdef['object_scale'])
loss.noobject_scale = float(mdef['noobject_scale'])
loss.class_scale = float(mdef['class_scale'])
loss.coord_scale = float(mdef['coord_scale'])
loss.input_size = (self.height, self.width)
modules = loss
elif mdef['type'] == 'globalmax':
modules = GlobalMaxPool2d()
elif mdef['type'] == 'reorg3d': # yolov3-spp-pan-scale
# torch.Size([16, 128, 104, 104])
# torch.Size([16, 64, 208, 208]) <-- # stride 2 interpolate dimensions 2 and 3 to cat with prior layer
pass
else:
print('Warning: Unrecognized Layer Type: ' + mdef['type'])
# Register module list and number of output filters
module_list.append(modules)
output_filters.append(filters)
return module_list, routs
def create_network(self, blocks):
models = nn.ModuleList()
prev_filters = 3
out_filters = []
conv_id = 0
dynamic_count = 0
for block in blocks:
if block['type'] == 'net' or block['type'] == 'learnet':
prev_filters = int(block['channels'])
continue
elif block['type'] == 'convolutional':
conv_id = conv_id + 1
batch_normalize = int(block['batch_normalize'])
filters = int(block['filters'])
kernel_size = int(block['size'])
stride = int(block['stride'])
is_pad = int(block['pad'])
# pad = (kernel_size-1)/2 if is_pad else 0 # for python2
pad = (kernel_size - 1) // 2 if is_pad else 0 # for python3
activation = block['activation']
groups = 1
bias = bool(int(block['bias'])) if 'bias' in block else True
if self.is_dynamic(block):
partial = int(
block['partial']) if 'partial' in block else None
Conv2d = dynamic_conv2d(dynamic_count == 0,
partial=partial)
dynamic_count += 1
else:
Conv2d = nn.Conv2d
if 'groups' in block:
groups = int(block['groups'])
model = nn.Sequential()
if batch_normalize:
model.add_module(
'conv{0}'.format(conv_id),
Conv2d(prev_filters,
filters,
kernel_size,
stride,
pad,
groups=groups,
bias=False))
model.add_module('bn{0}'.format(conv_id),
nn.BatchNorm2d(filters))
#model.add_module('bn{0}'.format(conv_id), BN2d(filters))
else:
model.add_module(
'conv{0}'.format(conv_id),
Conv2d(prev_filters,
filters,
kernel_size,
stride,
pad,
groups=groups,
bias=bias))
if activation == 'leaky':
model.add_module('leaky{0}'.format(conv_id),
nn.LeakyReLU(0.1, inplace=True))
elif activation == 'relu':
model.add_module('relu{0}'.format(conv_id),
nn.ReLU(inplace=True))
prev_filters = filters
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'maxpool':
pool_size = int(block['size'])
stride = int(block['stride'])
if stride > 1:
model = nn.MaxPool2d(pool_size, stride)
else:
model = MaxPoolStride1()
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'avgpool':
model = GlobalAvgPool2d()
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'softmax':
model = nn.Softmax()
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'cost':
if block['_type'] == 'sse':
model = nn.MSELoss(size_average=True)
elif block['_type'] == 'L1':
model = nn.L1Loss(size_average=True)
elif block['_type'] == 'smooth':
model = nn.SmoothL1Loss(size_average=True)
out_filters.append(1)
models.append(model)
elif block['type'] == 'reorg':
stride = int(block['stride'])
prev_filters = stride * stride * prev_filters
out_filters.append(prev_filters)
models.append(Reorg(stride))
elif block['type'] == 'route':
layers = block['layers'].split(',')
ind = len(models)
layers = [
int(i) if int(i) > 0 else int(i) + ind for i in layers
]
if len(layers) == 1:
prev_filters = out_filters[layers[0]]
elif len(layers) == 2:
assert (layers[0] == ind - 1)
prev_filters = out_filters[layers[0]] + out_filters[
layers[1]]
out_filters.append(prev_filters)
models.append(EmptyModule())
elif block['type'] == 'shortcut':
ind = len(models)
prev_filters = out_filters[ind - 1]
out_filters.append(prev_filters)
models.append(EmptyModule())
elif block['type'] == 'connected':
filters = int(block['output'])
if block['activation'] == 'linear':
model = nn.Linear(prev_filters, filters)
elif block['activation'] == 'leaky':
model = nn.Sequential(nn.Linear(prev_filters, filters),
nn.LeakyReLU(0.1, inplace=True))
elif block['activation'] == 'relu':
model = nn.Sequential(nn.Linear(prev_filters, filters),
nn.ReLU(inplace=True))
prev_filters = filters
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'region':
loss = RegionLossV2()
anchors = block['anchors'].split(',')
loss.anchors = [float(i) for i in anchors]
loss.num_classes = int(block['classes'])
loss.num_anchors = int(block['num'])
loss.anchor_step = len(loss.anchors) / loss.num_anchors
loss.object_scale = float(block['object_scale'])
loss.noobject_scale = float(block['noobject_scale'])
loss.class_scale = float(block['class_scale'])
loss.coord_scale = float(block['coord_scale'])
out_filters.append(prev_filters)
models.append(loss)
elif block['type'] == 'globalmax':
model = GlobalMaxPool2d()
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'globalavg':
model = GlobalAvgPool2d()
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'split':
splits = [int(sz) for sz in block['splits'].split(',')]
model = Split(splits)
prev_filters = splits[-1]
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'dynamic_routing':
model = DynamicRouting()
out_filters.append(prev_filters)
models.append(model)
else:
print('unknown type %s' % (block['type']))
# pdb.set_trace()
return models
def create_network(self, blocks):
models = nn.ModuleList()
prev_filters = 3
out_filters = []
conv_id = 0
dynamic_count = 0
for block in blocks:
if block['type'] == 'net' or block['type'] == 'learnet':
prev_filters = int(block['channels'])
continue
elif block['type'] == 'convolutional':
conv_id = conv_id + 1
batch_normalize = int(block['batch_normalize'])
filters = int(block['filters'])
kernel_size = int(block['size'])
stride = int(block['stride'])
is_pad = int(block['pad'])
pad = int((kernel_size - 1) / 2) if is_pad else 0
activation = block['activation']
groups = 1
bias = bool(int(block['bias'])) if 'bias' in block else True
if self.is_dynamic(block):
# don't know what partial parameter is doing, seems partial is always set to None
# partial = int(block['partial']) if 'partial' in block else None
# Conv2d = dynamic_conv2d(dynamic_count == 0, partial=partial)
# Conv2d = dynamic_conv2d(dynamic_count == 0)
Conv2d = EmbeddedConv
dynamic_count += 1
else:
Conv2d = nn.Conv2d
if 'groups' in block:
groups = int(block['groups'])
model = nn.Sequential()
if batch_normalize:
model.add_module(
'conv{0}'.format(conv_id),
Conv2d(prev_filters,
filters,
kernel_size,
stride,
pad,
groups=groups,
bias=False))
model.add_module('bn{0}'.format(conv_id),
nn.BatchNorm2d(filters))
else:
model.add_module(
'conv{0}'.format(conv_id),
Conv2d(prev_filters,
filters,
kernel_size,
stride,
pad,
groups=groups,
bias=bias))
if activation == 'leaky':
model.add_module('leaky{0}'.format(conv_id),
nn.LeakyReLU(0.1, inplace=True))
elif activation == 'relu':
model.add_module('relu{0}'.format(conv_id),
nn.ReLU(inplace=True))
prev_filters = filters
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'maxpool':
pool_size = int(block['size'])
stride = int(block['stride'])
# if stride > 1:
# model = nn.MaxPool2d(pool_size, stride)
# else:
# model = MaxPoolStride1()
model = nn.MaxPool2d(pool_size, stride)
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'globalavg':
model = nn.AdaptiveAvgPool2d(1)
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'globalmax':
model = nn.AdaptiveMaxPool2d(1)
out_filters.append(prev_filters)
models.append(model)
elif block['type'] == 'softmax':
model = nn.Softmax()
out_filters.append(prev_filters)
models.append(model)
# elif block['type'] == 'cost':
# if block['_type'] == 'sse':
# model = nn.MSELoss(size_average=True)
# elif block['_type'] == 'L1':
# model = nn.L1Loss(size_average=True)
# elif block['_type'] == 'smooth':
# model = nn.SmoothL1Loss(size_average=True)
# out_filters.append(1)
# models.append(model)
elif block['type'] == 'reorg':
stride = int(block['stride'])
prev_filters = stride * stride * prev_filters
out_filters.append(prev_filters)
models.append(Reorg(stride))
elif block['type'] == 'route':
layers = block['layers'].split(',')
ind = len(models)
layers = [
int(i) if int(i) > 0 else int(i) + ind for i in layers
]
if len(layers) == 1:
prev_filters = out_filters[layers[0]]
elif len(layers) == 2:
assert (layers[0] == ind - 1)
prev_filters = out_filters[layers[0]] + out_filters[
layers[1]]
out_filters.append(prev_filters)
models.append(EmptyModule())
elif block['type'] == 'shortcut':
ind = len(models)
prev_filters = out_filters[ind - 1]
out_filters.append(prev_filters)
models.append(EmptyModule())
elif block['type'] == 'connected':
filters = int(block['output'])
model = None
if block['activation'] == 'linear':
model = nn.Linear(prev_filters, filters)
elif block['activation'] == 'leaky':
model = nn.Sequential(nn.Linear(prev_filters, filters),
nn.LeakyReLU(0.1, inplace=True))
elif block['activation'] == 'relu':
model = nn.Sequential(nn.Linear(prev_filters, filters),
nn.ReLU(inplace=True))
prev_filters = filters
out_filters.append(prev_filters)
assert model is not None
models.append(model)
elif block['type'] == 'region':
anchors = block['anchors'].split(',')
anchors = [float(i) for i in anchors]
num_classes = int(block['classes'])
num_anchors = int(block['num'])
object_scale = float(block['object_scale'])
noobject_scale = float(block['noobject_scale'])
class_scale = float(block['class_scale'])
coord_scale = float(block['coord_scale'])
loss = RegionLossV2(num_classes, anchors, num_anchors,
coord_scale, noobject_scale, object_scale,
class_scale)
out_filters.append(prev_filters)
models.append(loss)
else:
print('unknown type %s' % (block['type']))
return models