def __init__(self, class_number=7):
super(Network_new, self).__init__()
self.class_number = class_number
self.extractor1 = nn.Sequential(
nn.Conv2d(in_channels=3,
out_channels=64,
kernel_size=3,
stride=1,
padding=1), RegionLayer_88(in_channels=64, grid=(4, 4)),
nn.ReLU(inplace=True), nn.MaxPool2d(kernel_size=2, stride=2))
self.extractor2 = nn.Sequential(
nn.Conv2d(in_channels=64,
out_channels=128,
kernel_size=3,
stride=1,
padding=1), RegionLayer_88(in_channels=128, grid=(4, 4)),
nn.ReLU(inplace=True), nn.MaxPool2d(kernel_size=2, stride=2))
self.extractor3 = nn.Sequential(
nn.Conv2d(in_channels=128,
out_channels=256,
kernel_size=3,
stride=1,
padding=1), RegionLayer_31(in_channels=256, grid=(4, 1)),
nn.ReLU(inplace=True), nn.MaxPool2d(kernel_size=2, stride=2))
self.bottleneck = nn.Sequential(
nn.Conv2d(256, 128, kernel_size=1, stride=1, padding=0,
bias=False), nn.BatchNorm2d(128), nn.ReLU(inplace=True),
nn.Conv2d(128, 128, 3, 1, 1, bias=False), nn.BatchNorm2d(128),
nn.ReLU(inplace=True),
nn.Conv2d(128, 256, kernel_size=1, stride=1, padding=0,
bias=False), nn.BatchNorm2d(256))
self.pool = nn.Sequential(nn.MaxPool2d(kernel_size=2, stride=2))
# self.classifier = nn.Sequential(
# nn.Linear(in_features=256, out_features=1024),
# nn.ReLU(inplace=True),
# nn.Dropout(0.2),
#
# nn.Linear(in_features=1024, out_features=1024),
# nn.ReLU(inplace=True),
# nn.Dropout(0.2),
#
# nn.Linear(in_features=1024, out_features=class_number)
# )
self.conv = nn.Conv2d(256, 256, 3, 2, 1)
self.avgpool = nn.AvgPool2d(kernel_size=6)
self.relu = nn.ReLU(inplace=True)
self.classifier = nn.Sequential(
nn.Linear(in_features=256, out_features=class_number))
def __init__(self, class_number=7):
super(Network_new, self).__init__()
self.class_number = class_number
self.extractor = nn.Sequential(
nn.Conv2d(in_channels=3,
out_channels=32,
kernel_size=3,
stride=1,
padding=1),
RegionLayer_88(in_channels=32, grid=(4, 4)),
# ReplaceRegionLayer(in_channels=32,),
nn.ReLU(),
nn.MaxPool2d(kernel_size=2, stride=2),
nn.BatchNorm2d(num_features=32),
nn.Conv2d(in_channels=32,
out_channels=16,
kernel_size=3,
stride=1,
padding=1),
nn.ReLU(),
nn.BatchNorm2d(num_features=16),
nn.Conv2d(in_channels=16,
out_channels=16,
kernel_size=3,
stride=1,
padding=1),
nn.ReLU(),
nn.MaxPool2d(kernel_size=2, stride=2),
nn.BatchNorm2d(num_features=16),
# nn.Conv2d(in_channels=16, out_channels=16, kernel_size=5, stride=2),
# nn.ReLU(),
nn.Conv2d(in_channels=16,
out_channels=16,
kernel_size=3,
stride=1,
padding=1),
RegionLayer_31(in_channels=16, grid=(4, 1)),
nn.ReLU(),
nn.MaxPool2d(kernel_size=2, stride=2),
nn.BatchNorm2d(num_features=16),
nn.Conv2d(in_channels=16,
out_channels=16,
kernel_size=3,
stride=1,
padding=1),
nn.ReLU(),
nn.BatchNorm2d(num_features=16),
nn.Conv2d(in_channels=16,
out_channels=16,
kernel_size=3,
stride=1,
padding=1),
nn.ReLU(),
)
self.classifier = nn.Sequential(
nn.Linear(in_features=16 * 6 * 6, out_features=4096), nn.ReLU(),
nn.Dropout(0.5), nn.Linear(in_features=4096, out_features=2048),
nn.ReLU(), nn.Dropout(0.5),
nn.Linear(in_features=2048, out_features=class_number))