yolov3学习笔记

yolov3

Yolov3整合当前网络上的一些优点,从下图可知，Yolov3的推理速度很快

参考github或代码

yolov3的map与推断时间与其他的对比

Top-1 Accuracy是指排名第一的类别与实际结果相符的准确率，
Top-5 Accuracy是指排名前五的类别包含实际结果的准确率。

yolov3的框架图

其主干网络采用的是Darknet-53

1. 为什么叫做Darknet-53:
因为他有53个卷积层
2. Darknet-53为什么要比ResNet-152要好？
答：1.darknet-53没有pool层，所有的下采样都是通过卷积层来实现(使用卷积层替换最大池化下采样)
3. Darknet-53为什么要比ResNet-152要快？
答：可能是因为他的卷积核要比ResNet-152个数要少

yolov3的3个特征图的输出大小

预测输出1尺寸为13×13，这里预测大的目标
预测输出3尺寸为52×52(细粒度最高)，用来预测小目标

yolov3的损失函数

置信度损失

使用logistic regression就是使用二值交叉熵损失

右边蓝色的代表的是anchor,绿色的是GT box,将预测边界框偏移量应用到anchor上得到黄色的目标边界框(预测目标边界框)
这里的黄色和绿色的IOU就是oi

类别损失

理解类别损失
这里的预测概率和不为1，因为这里使用的是二值交叉熵损失，每个预测值之间是相互独立的

定位损失

预测值是tx,ty,tw,th.

代码部分

代码部分的框架图如下:

代码阅读地址

yolo3-pytorch

阅读代码时发现github作者的实现好像跟论文还有点区别(比如论文的anchor框是由聚类产生的,但是作者是直接生成的)--后面注意要用到官方代码

Yolo3代码部分
step1:(nets目录下)
  darknet.py
step2:(nets目录下)
  yolo3.py
step3(utils目录下)
  utils.py
  对先验框调整的过程称为解码的过程
step4(根目录下)
  predict.py
  yolo.py
step5:(在根目录下/在utils目录下/在nets目录下)
  train.py
  dataloader.py
  yolo_training.py

print(model)产生的结果

用这个对着代码看能够更好的理解

YoloBody(
  (backbone): DarkNet(
    (conv1): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu1): LeakyReLU(negative_slope=0.1)
    (layer1): Sequential(
      (ds_conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (ds_bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (ds_relu): LeakyReLU(negative_slope=0.1)
      (residual_0): BasicBlock(
        (conv1): Conv2d(64, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(32, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
    )
    (layer2): Sequential(
      (ds_conv): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (ds_bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (ds_relu): LeakyReLU(negative_slope=0.1)
      (residual_0): BasicBlock(
        (conv1): Conv2d(128, 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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(64, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_1): BasicBlock(
        (conv1): Conv2d(128, 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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(64, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
    )
    (layer3): Sequential(
      (ds_conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (ds_bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (ds_relu): LeakyReLU(negative_slope=0.1)
      (residual_0): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_1): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_2): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_3): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_4): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_5): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_6): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_7): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(128, 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)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
    )
    (layer4): Sequential(
      (ds_conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (ds_bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (ds_relu): LeakyReLU(negative_slope=0.1)
      (residual_0): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_1): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_2): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_3): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_4): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_5): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_6): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_7): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
    )
    (layer5): Sequential(
      (ds_conv): Conv2d(512, 1024, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (ds_bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (ds_relu): LeakyReLU(negative_slope=0.1)
      (residual_0): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_1): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_2): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
      (residual_3): BasicBlock(
        (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)
        (relu1): LeakyReLU(negative_slope=0.1)
        (conv2): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (bn2): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (relu2): LeakyReLU(negative_slope=0.1)
      )
    )
  )
  (last_layer0): ModuleList(
    (0): Sequential(
      (conv): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (1): Sequential(
      (conv): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (2): Sequential(
      (conv): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (3): Sequential(
      (conv): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (4): Sequential(
      (conv): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (5): Sequential(
      (conv): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (6): Conv2d(1024, 18, kernel_size=(1, 1), stride=(1, 1))
  )
  (last_layer1_conv): Sequential(
    (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu): LeakyReLU(negative_slope=0.1)
  )
  (last_layer1_upsample): Upsample(scale_factor=2.0, mode=nearest)
  (last_layer1): ModuleList(
    (0): Sequential(
      (conv): Conv2d(768, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (1): Sequential(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (2): Sequential(
      (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (3): Sequential(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (4): Sequential(
      (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (5): Sequential(
      (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (6): Conv2d(512, 18, kernel_size=(1, 1), stride=(1, 1))
  )
  (last_layer2_conv): Sequential(
    (conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu): LeakyReLU(negative_slope=0.1)
  )
  (last_layer2_upsample): Upsample(scale_factor=2.0, mode=nearest)
  (last_layer2): ModuleList(
    (0): Sequential(
      (conv): Conv2d(384, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (1): Sequential(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (2): Sequential(
      (conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (3): Sequential(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (4): Sequential(
      (conv): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (5): Sequential(
      (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): LeakyReLU(negative_slope=0.1)
    )
    (6): Conv2d(256, 18, kernel_size=(1, 1), stride=(1, 1))
  )
)
Loading weights into state dict...

进程已结束，退出代码 -1

使用nertion生成的网络框架图

这个是用keras产生的(pytorch产生的时候线连不上--需要转换成onnx格式)

相关阅读:
c语言，回调函数[个人理解]
MFC、C++ 、Windows编程高手
 c++, 虚基派生 : 共同基类产生的二义性的解决办法
 c++，命名空间（namespace）
c++，纯虚函数与抽象类
 c++ ，protected 和 private修饰的构造函数
 c++ 虚析构函数[避免内存泄漏]
c++，虚函数
 c++，类的组合
 GPU与CPU的区别
原文地址：https://www.cnblogs.com/zranguai/p/14587152.html