Q&A
結果を切り取らないで全部載せたらいかがですか?
実現したいこと
あるコードで定義されているFaceEncoderモデルの構造を可視化する為、２つの方法を試しました。
① torchinfo.summary(model)
② print(model)
表示方法に違いはあれど、途中まで対応する層があるのですが、最後の方で①にはあるGatedConv2dとAdaptiveAvgPool2dの間のResNetと最後のConv2dが②にはありません。
こちらの理由を知りたいです！
よろしくお願いします。
① torchinfo.summary(model)
... ├─GatedConv2d: 1-7 [10, 1024, 16, 16] -- │ └─Sequential: 2-24 [10, 1024, 16, 16] -- │ │ └─Conv2d: 3-36 [10, 512, 16, 16] 4,719,104 │ │ └─BatchNorm2d: 3-37 [10, 512, 16, 16] 1,024 │ │ └─ReLU: 3-38 [10, 512, 16, 16] -- │ │ └─Conv2d: 3-39 [10, 1024, 16, 16] 4,719,616 │ │ └─Sigmoid: 3-40 [10, 1024, 16, 16] -- ├─ResNet: 1-8 -- (recursive) │ └─Sequential: 2-25 [10, 2048, 8, 8] -- │ │ └─Bottleneck: 3-41 [10, 2048, 8, 8] 6,039,552 │ │ └─Bottleneck: 3-42 [10, 2048, 8, 8] 4,462,592 │ │ └─Bottleneck: 3-43 [10, 2048, 8, 8] 4,462,592 ├─AdaptiveAvgPool2d: 1-9 [10, 2048, 1, 1] -- ├─Conv2d: 1-10 [10, 257, 1, 1] 526,593
② print(model)
... (gated): GatedConv2d( (ConvMask): Sequential( (0): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (2): ReLU(inplace=True) (3): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (4): Sigmoid() ) ) (avgpool): AdaptiveAvgPool2d(output_size=(1, 1)) )
該当のソースコード
python network.pyで①、②を可視化。
import torch from torch import nn # from models.resnet50 import resnet50 from resnet50 import resnet50 import torch.nn.functional as F from torchinfo import summary class ConvBNReLU(nn.Module): def __init__(self, in_chan, out_chan, ks=3, stride=1, padding=1, *args, **kwargs): super(ConvBNReLU, self).__init__() self.conv = nn.Conv2d(in_chan, out_chan, kernel_size=ks, stride=stride, padding=padding, bias=False) self.bn = nn.BatchNorm2d(out_chan) self.init_weight() def init_weight(self): for ly in self.children(): if isinstance(ly, nn.Conv2d): nn.init.kaiming_normal_(ly.weight, a=1) if not ly.bias is None: nn.init.constant_(ly.bias, 0) def forward(self, x): x = self.conv(x) x = F.relu(self.bn(x)) return x class AttentionRefinementModule(nn.Module): def __init__(self, in_chan, out_chan): super().__init__() self.conv = ConvBNReLU(in_chan, out_chan, ks=3, stride=1, padding=1) self.conv_atten = nn.Conv2d(out_chan, out_chan, kernel_size=1, bias=False) self.bn_atten = nn.BatchNorm2d(out_chan) self.sigmoid_atten = nn.Sigmoid() self.init_weight() def init_weight(self): for ly in self.children(): if isinstance(ly, nn.Conv2d): nn.init.kaiming_normal_(ly.weight, a=1) if not ly.bias is None: nn.init.constant_(ly.bias, 0) def forward(self, x): feat = self.conv(x) atten = F.avg_pool2d(feat, feat.size()[2:]) atten = self.conv_atten(atten) atten = self.bn_atten(atten) atten = self.sigmoid_atten(atten) out = torch.mul(feat, atten) return out class FeatureFusionModule(nn.Module): def __init__(self, in_chan, out_chan): super().__init__() self.convblk = ConvBNReLU(in_chan, out_chan, ks=1, stride=1, padding=0) self.conv1 = nn.Conv2d(out_chan, out_chan // 4, kernel_size=1, stride=1, padding=0, bias=False) self.conv2 = nn.Conv2d(out_chan // 4, out_chan, kernel_size=1, stride=1, padding=0, bias=False) self.relu = nn.ReLU(inplace=True) self.sigmoid = nn.Sigmoid() self.init_weight() def init_weight(self): for ly in self.children(): if isinstance(ly, nn.Conv2d): nn.init.kaiming_normal_(ly.weight, a=1) if not ly.bias is None: nn.init.constant_(ly.bias, 0) def forward(self, fsp, fcp): fcat = torch.cat([fsp, fcp], dim=1) feat = self.convblk(fcat) atten = F.avg_pool2d(feat, feat.size()[2:]) atten = self.conv1(atten) atten = self.relu(atten) atten = self.conv2(atten) atten = self.sigmoid(atten) feat_atten = torch.mul(feat, atten) feat_out = feat_atten + feat return feat_out class ContextPath(nn.Module): def __init__(self): super().__init__() self.arm8 = AttentionRefinementModule(512, 256) self.arm16 = AttentionRefinementModule(1024, 256) self.conv_head8 = ConvBNReLU(256, 256, 3, 1, 1) self.conv_head16 = ConvBNReLU(256, 256, 3, 1, 1) self.conv_avg = ConvBNReLU(1024, 256, 1, 1, 0) self.init_weight() def init_weight(self): for ly in self.children(): if isinstance(ly, nn.Conv2d): nn.init.kaiming_normal_(ly.weight, a=1) if not ly.bias is None: nn.init.constant_(ly.bias, 0) def forward(self, feat4, feat8, feat16): H4, W4 = feat4.shape[2:] H8, W8 = feat8.shape[2:] H16, W16 = feat16.shape[2:] avg = F.avg_pool2d(feat16, feat16.size()[2:]) avg = self.conv_avg(avg) avg_up = F.interpolate(avg, (H16, W16), mode='nearest') feat16_arm = self.arm16(feat16) feat16_sum = feat16_arm + avg_up feat16_up = F.interpolate(feat16_sum, (H8, W8), mode='nearest') feat16_up = self.conv_head16(feat16_up) feat8_arm = self.arm8(feat8) feat8_sum = feat8_arm + feat16_up feat8_up = F.interpolate(feat8_sum, (H4, W4), mode='nearest') feat8_up = self.conv_head8(feat8_up) return feat4, feat8_up, feat16_up class BiSeNetOutPut(nn.Module): def __init__(self, in_chan, mid_chan, n_classes): super().__init__() self.conv = ConvBNReLU(in_chan, mid_chan, ks=3, stride=1, padding=1) self.conv_out = nn.Conv2d(mid_chan, n_classes, kernel_size=1, bias=False) self.init_weight() def init_weight(self): for ly in self.children(): if isinstance(ly, nn.Conv2d): nn.init.kaiming_normal_(ly.weight, a=1) if not ly.bias is None: nn.init.constant_(ly.bias, 0) def forward(self, x): x = self.conv(x) x = self.conv_out(x) return x class GatedConv2d(nn.Module): def __init__(self, in_chan, mid_chan): super().__init__() self.ConvMask = nn.Sequential( nn.Conv2d(in_chan, mid_chan, 3, 1, 1), nn.BatchNorm2d(mid_chan), nn.ReLU(inplace=True), nn.Conv2d(mid_chan, in_chan, 3, 1, 1), nn.Sigmoid() ) self.init_weight() def init_weight(self): for ly in self.children(): if isinstance(ly, nn.Conv2d): nn.init.kaiming_normal_(ly.weight, a=1) if not ly.bias is None: nn.init.constant_(ly.bias, 0) def forward(self, feat): mask = self.ConvMask(feat) out = feat * mask return out class FaceEncoder(nn.Module): def __init__(self): super(FaceEncoder, self).__init__() self.feat = resnet50(pretrained=True) # 事前学習済みのResNet-50モデルを使用して特徴マップを抽出するための部分モジュール self.coeff = nn.Conv2d(2048, 257, (1, 1)) self.cp = ContextPath() self.ffm = FeatureFusionModule(512, 256) self.conv_out = BiSeNetOutPut(256, 256, n_classes=1) self.conv_out8 = BiSeNetOutPut(256, 128, n_classes=1) self.conv_out16 = BiSeNetOutPut(256, 128, n_classes=1) self.gated = GatedConv2d(1024, 512) self.avgpool = nn.AdaptiveAvgPool2d((1, 1)) self.weight_init() @torch.no_grad() def weight_init(self): self.coeff.weight.zero_() self.coeff.bias.zero_() def forward(self, img): H, W = img.shape[2:] # 入力画像から特徴マップを抽出（feat4, feat8, feat16） feat4, feat8, feat16 = self.feat(img) # 1/4, 1/8. 1/16 of initial size : [10, 256, 64, 64], [10, 512, 32, 32], [10, 1024, 16, 16] feat_res4, feat_cp4, feat_cp8 = self.cp(feat4, feat8, feat16) # [10, 256, 64, 64], [10, 256, 64, 64], [10, 256, 32, 32] feat_sp = feat_res4 feat_fuse = self.ffm(feat_sp, feat_cp4) feat_out = self.conv_out(feat_fuse) feat_out8 = self.conv_out8(feat_cp4) feat_out16 = self.conv_out16(feat_cp8) feat_out = F.interpolate(feat_out, (H, W), mode='bilinear', align_corners=True) feat_out8 = F.interpolate(feat_out8, (H, W), mode='bilinear', align_corners=True) feat_out16 = F.interpolate(feat_out16, (H, W), mode='bilinear', align_corners=True) feat16 = self.gated(feat16) feat = self.feat.layer4(feat16) # [10, 2048, 8, 8] feat = self.avgpool(feat) # [10, 2048, 1, 1] coeff = self.coeff(feat) # [10, 257, 1, 1] return coeff.squeeze(2).squeeze(2), feat_out, feat_out8, feat_out16 if __name__ == '__main__': model = FaceEncoder() print(model) summary(model=FaceEncoder(), input_size=(10, 3, 256, 256))
@quickquipさんご指摘ありがとうございます。
以下に結果の全体像を載せます。
↓ torchinfo.summary
===============================================================================================
Layer (type:depth-idx) Output Shape Param #
===============================================================================================
FaceEncoder [10, 257] --
├─ResNet: 1-1 [10, 256, 64, 64] 14,964,736
│ └─Conv2d: 2-1 [10, 64, 128, 128] 9,408
│ └─BatchNorm2d: 2-2 [10, 64, 128, 128] 128
│ └─ReLU: 2-3 [10, 64, 128, 128] --
│ └─MaxPool2d: 2-4 [10, 64, 64, 64] --
│ └─Sequential: 2-5 [10, 256, 64, 64] --
│ │ └─Bottleneck: 3-1 [10, 256, 64, 64] 75,008
│ │ └─Bottleneck: 3-2 [10, 256, 64, 64] 70,400
│ │ └─Bottleneck: 3-3 [10, 256, 64, 64] 70,400
│ └─Sequential: 2-6 [10, 512, 32, 32] --
│ │ └─Bottleneck: 3-4 [10, 512, 32, 32] 379,392
│ │ └─Bottleneck: 3-5 [10, 512, 32, 32] 280,064
│ │ └─Bottleneck: 3-6 [10, 512, 32, 32] 280,064
│ │ └─Bottleneck: 3-7 [10, 512, 32, 32] 280,064
│ └─Sequential: 2-7 [10, 1024, 16, 16] --
│ │ └─Bottleneck: 3-8 [10, 1024, 16, 16] 1,512,448
│ │ └─Bottleneck: 3-9 [10, 1024, 16, 16] 1,117,184
│ │ └─Bottleneck: 3-10 [10, 1024, 16, 16] 1,117,184
│ │ └─Bottleneck: 3-11 [10, 1024, 16, 16] 1,117,184
│ │ └─Bottleneck: 3-12 [10, 1024, 16, 16] 1,117,184
│ │ └─Bottleneck: 3-13 [10, 1024, 16, 16] 1,117,184
├─ContextPath: 1-2 [10, 256, 64, 64] --
│ └─ConvBNReLU: 2-8 [10, 256, 1, 1] --
│ │ └─Conv2d: 3-14 [10, 256, 1, 1] 262,144
│ │ └─BatchNorm2d: 3-15 [10, 256, 1, 1] 512
│ └─AttentionRefinementModule: 2-9 [10, 256, 16, 16] --
│ │ └─ConvBNReLU: 3-16 [10, 256, 16, 16] 2,359,808
│ │ └─Conv2d: 3-17 [10, 256, 1, 1] 65,536
│ │ └─BatchNorm2d: 3-18 [10, 256, 1, 1] 512
│ │ └─Sigmoid: 3-19 [10, 256, 1, 1] --
│ └─ConvBNReLU: 2-10 [10, 256, 32, 32] --
│ │ └─Conv2d: 3-20 [10, 256, 32, 32] 589,824
│ │ └─BatchNorm2d: 3-21 [10, 256, 32, 32] 512
│ └─AttentionRefinementModule: 2-11 [10, 256, 32, 32] --
│ │ └─ConvBNReLU: 3-22 [10, 256, 32, 32] 1,180,160
│ │ └─Conv2d: 3-23 [10, 256, 1, 1] 65,536
│ │ └─BatchNorm2d: 3-24 [10, 256, 1, 1] 512
│ │ └─Sigmoid: 3-25 [10, 256, 1, 1] --
│ └─ConvBNReLU: 2-12 [10, 256, 64, 64] --
│ │ └─Conv2d: 3-26 [10, 256, 64, 64] 589,824
│ │ └─BatchNorm2d: 3-27 [10, 256, 64, 64] 512
├─FeatureFusionModule: 1-3 [10, 256, 64, 64] --
│ └─ConvBNReLU: 2-13 [10, 256, 64, 64] --
│ │ └─Conv2d: 3-28 [10, 256, 64, 64] 131,072
│ │ └─BatchNorm2d: 3-29 [10, 256, 64, 64] 512
│ └─Conv2d: 2-14 [10, 64, 1, 1] 16,384
│ └─ReLU: 2-15 [10, 64, 1, 1] --
│ └─Conv2d: 2-16 [10, 256, 1, 1] 16,384
│ └─Sigmoid: 2-17 [10, 256, 1, 1] --
├─BiSeNetOutPut: 1-4 [10, 1, 64, 64] --
│ └─ConvBNReLU: 2-18 [10, 256, 64, 64] --
│ │ └─Conv2d: 3-30 [10, 256, 64, 64] 589,824
│ │ └─BatchNorm2d: 3-31 [10, 256, 64, 64] 512
│ └─Conv2d: 2-19 [10, 1, 64, 64] 256
├─BiSeNetOutPut: 1-5 [10, 1, 64, 64] --
│ └─ConvBNReLU: 2-20 [10, 128, 64, 64] --
│ │ └─Conv2d: 3-32 [10, 128, 64, 64] 294,912
│ │ └─BatchNorm2d: 3-33 [10, 128, 64, 64] 256
│ └─Conv2d: 2-21 [10, 1, 64, 64] 128
├─BiSeNetOutPut: 1-6 [10, 1, 32, 32] --
│ └─ConvBNReLU: 2-22 [10, 128, 32, 32] --
│ │ └─Conv2d: 3-34 [10, 128, 32, 32] 294,912
│ │ └─BatchNorm2d: 3-35 [10, 128, 32, 32] 256
│ └─Conv2d: 2-23 [10, 1, 32, 32] 128
├─GatedConv2d: 1-7 [10, 1024, 16, 16] --
│ └─Sequential: 2-24 [10, 1024, 16, 16] --
│ │ └─Conv2d: 3-36 [10, 512, 16, 16] 4,719,104
│ │ └─BatchNorm2d: 3-37 [10, 512, 16, 16] 1,024
│ │ └─ReLU: 3-38 [10, 512, 16, 16] --
│ │ └─Conv2d: 3-39 [10, 1024, 16, 16] 4,719,616
│ │ └─Sigmoid: 3-40 [10, 1024, 16, 16] --
├─ResNet: 1-8 -- (recursive)
│ └─Sequential: 2-25 [10, 2048, 8, 8] --
│ │ └─Bottleneck: 3-41 [10, 2048, 8, 8] 6,039,552
│ │ └─Bottleneck: 3-42 [10, 2048, 8, 8] 4,462,592
│ │ └─Bottleneck: 3-43 [10, 2048, 8, 8] 4,462,592
├─AdaptiveAvgPool2d: 1-9 [10, 2048, 1, 1] --
├─Conv2d: 1-10 [10, 257, 1, 1] 526,593
===============================================================================================
Total params: 54,900,033
Trainable params: 54,900,033
Non-trainable params: 0
Total mult-adds (G): 170.52
===============================================================================================
Input size (MB): 7.86
Forward/backward pass size (MB): 3067.99
Params size (MB): 159.74
Estimated Total Size (MB): 3235.60
===============================================================================================
↓ print(model)
FaceEncoder(
(feat): ResNet(
(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
(layer1): Sequential(
(0): Bottleneck(
(conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
)
(layer2): Sequential(
(0): Bottleneck(
(conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(3): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
)
(layer3): Sequential(
(0): Bottleneck(
(conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(3): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(4): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
(5): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
)
)
(layer4): Sequential(
(0): Bottleneck(
(conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(2048, 512, kernel_si
よろしくお願いいたします。。
質問は編集できます。情報はこの欄ではなくて質問に載せましょう。
回答1件
0
自己解決
・print(model)は単にinitに記載したモデル構造を表示
・torchinfo.summaryはforwardに沿ってデータの流れを可視化
今回の場合は、再帰的にfeatのlayer4を下層で使用していた為、print(model)では見当たらなかった
投稿2023/07/09 14:01
総合スコア25
あなたの回答
tips
プレビュー
