def __init__(self, block=Bottleneck, layers=[3, 4, 6, 3], nb_classes=40, channel=3): self.inplanes = 64 super(ResNet, self).__init__() self.conv1_custom = nn.Conv2d(channel, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1 = nn.BatchNorm2d(64) self.relu = nn.ReLU(inplace=True) self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) self.layer1 = self._make_layer(block, 64, layers[0]) self.layer2 = self._make_layer(block, 128, layers[1], stride=2) self.layer3 = self._make_layer(block, 256, layers[2], stride=2) self.layer4 = self._make_layer(block, 512, layers[3], stride=2) self.avgpool = nn.AvgPool2d(2) #self.maxpool = nn.MaxPool2d(kernel_size=7,stride=2) #self.channelpool = nn.AvgPool1d(1) #self.bn_global = nn.BatchNorm1d(2048) #self.fc_custom = nn.Linear(512 * block.expansion, nb_classes) #self.fc_pool = nn.Conv1d(50,1,kernel_size=1,stride=1) #self.fc_s = nn.Linear(2048,nb_classes) self.extractor = SingleRoIExtractor(dict(type='RoIAlign', out_size=7, sample_num=1), out_channels=512, featmap_strides=[ 4, ]) self.attention = MultiHeadAttention(1, 2048, 512, 512) self.fc_roi = nn.Linear(2048 * 16, 40) self.bn2 = nn.BatchNorm2d(2048) #self.fc_global = nn.Linear(2048*7*7,2048) #self.fc_s2 = nn.Linear(1024,nb_classes) #self.conv_end = nn.Conv2d(1,1,kernel_size=(16,1),stride=1) 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)) elif isinstance(m, nn.BatchNorm2d): m.weight.data.fill_(1) m.bias.data.zero_()
def __init__(self, block, layers, num_classes=40): self.inplanes = 64 super(ResNet, self).__init__() self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1 = nn.BatchNorm2d(64) self.relu = nn.ReLU(inplace=True) self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) self.layer1 = self._make_layer(block, 64, layers[0]) self.layer2 = self._make_layer(block, 128, layers[1], stride=2) self.layer3 = self._make_layer(block, 256, layers[2], stride=2) self.layer4 = self._make_layer(block, 512, layers[3], stride=2) self.conv1g = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1g = nn.BatchNorm2d(64) self.relug = nn.ReLU(inplace=True) self.maxpoolg = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) self.layer1g = self.layer1 self.layer2g = self.layer2 self.layer3g = self.layer3 self.layer4g = self.layer4 self.avgpooll = nn.AvgPool2d(2) self.avgpoolg = nn.AvgPool2d(7) #self.fcadd = nn.Linear(2048,2048) #self.atn_s1 = MultiHeadAttention(56*56,256,64,64) #self.atn_s2 = MultiHeadAttention(28*28,512,128,128) self.atn_s4 = MultiHeadAttention(1,2048*2*2,512,512) #self.atn_s3 = MultiHeadAttention(14*14,1024,256,256) # self.fc_aux = nn.Linear(512 * block.expansion, 101) self.dp = nn.Dropout(p=0.8) self.fc_g = nn.Linear(512 * block.expansion, num_classes) self.fc_l = nn.Linear(512 * block.expansion, num_classes) # self.bn_final = nn.BatchNorm1d(num_classes) # self.fc2 = nn.Linear(num_classes, num_classes) # self.fc_final = nn.Linear(num_classes, 101) self.extractor = SingleRoIExtractor(dict(type='RoIAlign', out_size=2, sample_num=1),out_channels=2048, featmap_strides=[32, ] ) 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)) elif isinstance(m, nn.BatchNorm2d): m.weight.data.fill_(1) m.bias.data.zero_()
def __init__(self): super(proposalModel, self).__init__() self.backbone = ResNet(101, 4, out_indices=( 1, 3, )) # int obj不可迭代,加个 , self.roiextract = SingleRoIExtractor(roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2), out_channels=512, featmap_strides=[ 8, ]) # 没有FPN 所以只有一个步长 self.fc = nn.Linear(2048 + 2048, 40) self.backbone.init_weights( pretrained='/home/share/LabServer/GLnet/MODELZOO/resnet50.pth') self.conv1 = nn.Conv2d(512, 1024, 3, 1) self.conv2 = nn.Conv2d(1024, 2048, 3, 1)
def __init__(self, roi_out_size=7, roi_sample_num=2, channels=256, strides=[4, 8, 16, 32], featmap_num=5): super(RPN_Modulator, self).__init__() self.roi_extractor = SingleRoIExtractor(roi_layer={ 'type': 'RoIAlign', 'out_size': roi_out_size, 'sample_num': roi_sample_num }, out_channels=channels, featmap_strides=strides) self.proj_modulator = nn.ModuleList([ nn.Conv2d(channels, channels, roi_out_size, padding=0) for _ in range(featmap_num) ]) self.proj_out = nn.ModuleList([ nn.Conv2d(channels, channels, 1, padding=0) for _ in range(featmap_num) ])
def __init__(self): super(proposalModel_scene, self).__init__() self.HIGHSCENE = [ 3, 4, 6, 7, 9, 10, 13, 12, 11, 16, 24, 25, 26, 32, 39 ] self.backbone = ResNet(50, 4, out_indices=( 1, 3, )) # int obj不可迭代,加个 , self.roiextract = SingleRoIExtractor(roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2), out_channels=512, featmap_strides=[ 8, ]) # 没有FPN 所以只有一个步长 self.fc_high = nn.Linear(2048 + 1024, 40) self.fc_low = nn.Linear(1024 + 1024, 40) self.conv1 = nn.Conv2d(512, 1024, 3, 1, 0) self.conv2 = nn.Conv2d(1024, 2048, 3, 1, 0) self.backbone.init_weights( pretrained='/home/share/LabServer/GLnet/MODELZOO/resnet50.pth')
def __init__(self, block=Bottleneck, layers=[3, 4, 6, 3], nb_classes=40, channel=3): self.inplanes = 64 super(ResNet, self).__init__() self.conv1_custom = nn.Conv2d(channel, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1 = nn.BatchNorm2d(64) self.relu = nn.ReLU(inplace=True) self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) self.layer1 = self._make_layer(block, 64, layers[0]) self.layer2 = self._make_layer(block, 128, layers[1], stride=2) self.layer3 = self._make_layer(block, 256, layers[2], stride=2) self.layer4 = self._make_layer(block, 512, layers[3], stride=2) self.avgpool = nn.AvgPool2d(7) self.maxpool = nn.MaxPool2d(kernel_size=7, stride=2) self.channelpool = nn.AvgPool1d(100) #self.fc_custom = nn.Linear(512 * block.expansion, nb_classes) #self.fc_at1 = nn.Linear(1024,nb_classes) #self.fc_at2 = nn.Linear(2048,nb_classes) """ self.conv_lower = nn.Sequential(OrderedDict([ ('con1',nn.Conv2d(1024,521,kernel_size=3,stride=2,padding=1)), ('bn1',nn.BatchNorm2d(521)), ('relu1',nn.ReLU()), ('conv2',(nn.Conv2d(521,256,kernel_size=3,stride=1,padding=1))), ('bn2',nn.BatchNorm2d(256)), ('relu2',nn.ReLU()), ('conv3',(nn.Conv2d(256,10,kernel_size=1,stride=1,padding=0))), ('bn3',nn.BatchNorm2d(10)), ('relu3',nn.ReLU()), ('conv4',(nn.Conv2d(10,1,kernel_size=1,stride=1))), ('relu4',nn.ReLU()), ])) """ #self.conv_atten = self.layer4 #self.layer3_main = self.layer3 #self.layer4_main = self.layer4 self.fc_custom = nn.Linear(2048 * 2, nb_classes) self.extractor = SingleRoIExtractor(dict(type='RoIAlign', out_size=3, sample_num=2), out_channels=1024, featmap_strides=[ 16, ]) self.attention_op = MultiHeadAttention(1, 2048, 2048, 2048) self.convx = nn.Conv2d(1024, 2048, kernel_size=3, stride=2) 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)) elif isinstance(m, nn.BatchNorm2d): m.weight.data.fill_(1) m.bias.data.zero_()