基于ResNet-152模型的172种食物图像分类（教育部产学合作协同育人项目）-人工智能-PHP中文网

本项目基于paddlepaddle和VireoFood-172数据集，用ResNet模型分三段训练实现食品识别。经不同学习率多轮训练，ResNet152模型训练集正确率92.927%，测试集82.684%，后续将优化以提升精度。

☞☞☞AI 智能聊天, 问答助手, AI 智能搜索, 免费无限量使用 DeepSeek R1 模型☜☜☜

基于resnet-152模型的172种食物图像分类（教育部产学合作协同育人项目） - php中文网

一、项目背景和项目创意

由于计算机视觉技术在监控摄像头、医疗保健等许多领域的应用越来越多。食品识别是其中一个重要的领域，由于其现实意义和科学挑战，值得进一步研究。
最近，卷积神经网络(CNN)被用于食品识别。食物识别方法使用CNN模型提取食物图像特征，计算食物图像特征的相似度，并使用分类技术训练分类器来完成食物识别。
此项目利用了VireoFood-172 数据集，其中包含了来自172个类别的110241张食品图片，并根据353种配料手工标注，以此来进行训练和测试。

二、项目搭建

Step1:准备数据

（1）数据集介绍

本项目基于paddlepaddle，利用VireoFood-172 数据集进行相关开发，此数据集将食品分为172个大类，每大类中有200-1k张从从百度和谷歌图像搜索中抓取的食品图片，基本覆盖日常生活中的绝大多数食品种类。数据集数量庞大，质量较高，能够满足深度学习的训练要求。

（2）对数据集进行解压

In [ ]

!unzip -q -o data/data124758/splitedDataset.zip

登录后复制

（3）对数据集进行处理

拿到数据集之后首先对数据集进行处理，按8:2的比例切割数据集之后进行随机打乱

（4）引入项目必须的模块

In [2]

import paddleimport numpy as npimport os

登录后复制

（5）利用生成器获得训练集和测试集

引入自己编写的生成器代码FoodDataset
训练集：train_dataset = FoodDataset(train_data)
测试集：eval_dataset = FoodDataset(validation_data)
train大小： 88192
eval大小： 22049
符合按8:2的比例切割数据集

In [3]

from dataset import FoodDataset

登录后复制

In [4]

train_data = './train_data1.txt'validation_data = './valida_data1.txt'train_dataset = FoodDataset(train_data)
eval_dataset = FoodDataset(validation_data)print('train大小：', train_dataset.__len__())print('eval大小：', eval_dataset.__len__())# for data, label in train_dataset:#     print(data)#     print(np.array(data).shape)#     print(label)#     break

登录后复制

train大小： 88192
eval大小： 22049

登录后复制

In [ ]

class DenseFoodModel(paddle.nn.Layer):
    def __init__(self):
        super(DenseFoodModel, self).__init__()
        self.num_labels = 172

    def dense_block(self, x, blocks, name):
        for i in range(blocks):
            x = self.conv_block(x, 32, name=name + '_block' + str(i + 1))        return x    def transition_block(self, x, reduction, name):
        bn_axis = 3
        x = paddle.nn.BatchNorm2D(num_features=x.shape[1], epsilon=1.001e-5)(x)
        x = paddle.nn.ELU(name=name + '_elu')(x)
        x = paddle.nn.Conv2D(in_channels=x.shape[1], out_channels=int(x.shape[1] * reduction),
                             kernel_size=1, bias_attr=False)(x)
        x = paddle.nn.MaxPool2D(kernel_size=2, stride=2, name=name + '_pool')(x)        return x    def conv_block(self, x, growth_rate, name):
        bn_axis = 3
        x1 = paddle.nn.BatchNorm2D(num_features=x.shape[1], epsilon=1.001e-5)(x)
        x1 = paddle.nn.ELU(name=name + '_0_elu')(x1)
        x1 = paddle.nn.Conv2D(in_channels=x1.shape[1], out_channels=4 * growth_rate, kernel_size= 1, bias_attr=False)(x1)
        x1 = paddle.nn.BatchNorm2D(num_features=x1.shape[1], epsilon=1.001e-5)(x1)
        x1 = paddle.nn.ELU(name=name + '_1_elu')(x1)
        x1 = paddle.nn.Conv2D(in_channels=x1.shape[1], out_channels=growth_rate, kernel_size=3, padding='SAME',
                              bias_attr=False)(x1)        # x = np.concatenate(([x, x1]), axis=1)
        # x = paddle.to_tensor(x)
        x = paddle.concat(x=[x, x1], axis=1)        return x    def forward(self, input):
        # img_input = paddle.reshape(input, shape=[-1, 3, 224, 224])  # 转换维读
        bn_axis = 3
        x = paddle.nn.Conv2D(in_channels=3, out_channels=64, kernel_size=7, stride=2, bias_attr=False, padding=3)(input)
        x = paddle.nn.BatchNorm2D(
            num_features=64, epsilon=1.001e-5)(x)
        x = paddle.nn.ELU(name='conv1/elu')(x)
        x = paddle.nn.MaxPool2D(kernel_size=3, stride=2, name='pool1', padding=1)(x)
        x = self.dense_block(x, 6, name='conv2')
        x = self.transition_block(x, 0.5, name='pool2')
        x = self.dense_block(x, 12, name='conv3')
        x = self.transition_block(x, 0.5, name='pool3')
        x = self.dense_block(x, 24, name='conv4')
        x = self.transition_block(x, 0.5, name='pool4')
        x = self.dense_block(x, 16, name='conv5')
        x = paddle.nn.BatchNorm2D(
            num_features=x.shape[1], epsilon=1.001e-5)(x)
        x = paddle.nn.ELU(name='elu')(x)
        x = paddle.nn.AdaptiveAvgPool2D(output_size=1)(x)
        x = paddle.squeeze(x, axis=[2, 3])
        x = paddle.nn.Linear(in_features=1024, out_features=173)(x)
        x = F.softmax(x)        return x

登录后复制

In [ ]

import paddle.nn.functional as F

model = paddle.Model(DenseFoodModel())
model.summary((-1, ) + tuple([3, 224, 224]))

登录后复制

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/nn/layer/norm.py:653: UserWarning: When training, we now always track global mean and variance.
  "When training, we now always track global mean and variance.")

登录后复制

----------------------------------------------------------------------------
  Layer (type)       Input Shape          Output Shape         Param #    
============================================================================
DenseFoodModel-1  [[1, 3, 224, 224]]        [1, 173]              0       
============================================================================
Total params: 0
Trainable params: 0
Non-trainable params: 0
----------------------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 0.00
Params size (MB): 0.00
Estimated Total Size (MB): 0.58
----------------------------------------------------------------------------

登录后复制

{'total_params': 0, 'trainable_params': 0}

登录后复制

Step2.网络配置

（1）网络搭建

此项目采用paddlepaddle自带的ResNet残差网络模型，ResNet残差模型快如下图所示：

基于ResNet-152模型的172种食物图像分类（教育部产学合作协同育人项目） - php中文网

x为残差块的输入，然后复制成两部分，一部分输入到层（weight layer）之中，进行层间的运算（相当于将x输入到一个函数中做映射）,结果为f(x);另一部分作为分支结构，输出还是原本的x，最后将分别两部分的输出进行叠加:f(x) + x，再通过激活函数。这便是整个残差块的基本结构。
下图是每种ResNet的具体结构：

基于ResNet-152模型的172种食物图像分类（教育部产学合作协同育人项目） - php中文网

这里介绍一下ResNet152，152是指152次卷积。
其中block共有3+8+36+3 = 50个，每个block是由3层卷积构成的，共150个卷积，最开始的一个卷积是将3通道的图片提取特征，最后一层是自适应平均池化，输出维度为1。
一开始选用的是ResNet50图像分类模型，但是在进行了100多轮迭代训练后，正确率只能维持在88%左右，因此我们尝试变换模型重新进行训练，选用了ResNet101模型和ResNet152模型分别进行100轮迭代训练，最终，ResNet101模型的训练集正确率88%，测试集正确率80%，ResNet152模型训练集正确率达到了92%，测试集正确率达到了82%。

In [5]

network = paddle.vision.models.resnet152(num_classes=173, pretrained=True)
model = paddle.Model(network)
model.summary((-1, ) + tuple([3, 224, 224]))

登录后复制

100%|██████████| 355826/355826 [00:09<00:00, 38920.04it/s]
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py:1441: UserWarning: Skip loading for fc.weight. fc.weight receives a shape [2048, 1000], but the expected shape is [2048, 173].
  warnings.warn(("Skip loading for {}. ".format(key) + str(err)))
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py:1441: UserWarning: Skip loading for fc.bias. fc.bias receives a shape [1000], but the expected shape is [173].
  warnings.warn(("Skip loading for {}. ".format(key) + str(err)))

登录后复制

-------------------------------------------------------------------------------
   Layer (type)         Input Shape          Output Shape         Param #    
===============================================================================
     Conv2D-1        [[1, 3, 224, 224]]   [1, 64, 112, 112]        9,408     
   BatchNorm2D-1    [[1, 64, 112, 112]]   [1, 64, 112, 112]         256      
      ReLU-1        [[1, 64, 112, 112]]   [1, 64, 112, 112]          0       
    MaxPool2D-1     [[1, 64, 112, 112]]    [1, 64, 56, 56]           0       
     Conv2D-3        [[1, 64, 56, 56]]     [1, 64, 56, 56]         4,096     
   BatchNorm2D-3     [[1, 64, 56, 56]]     [1, 64, 56, 56]          256      
      ReLU-2         [[1, 256, 56, 56]]    [1, 256, 56, 56]          0       
     Conv2D-4        [[1, 64, 56, 56]]     [1, 64, 56, 56]        36,864     
   BatchNorm2D-4     [[1, 64, 56, 56]]     [1, 64, 56, 56]          256      
     Conv2D-5        [[1, 64, 56, 56]]     [1, 256, 56, 56]       16,384     
   BatchNorm2D-5     [[1, 256, 56, 56]]    [1, 256, 56, 56]        1,024     
     Conv2D-2        [[1, 64, 56, 56]]     [1, 256, 56, 56]       16,384     
   BatchNorm2D-2     [[1, 256, 56, 56]]    [1, 256, 56, 56]        1,024     
 BottleneckBlock-1   [[1, 64, 56, 56]]     [1, 256, 56, 56]          0       
     Conv2D-6        [[1, 256, 56, 56]]    [1, 64, 56, 56]        16,384     
   BatchNorm2D-6     [[1, 64, 56, 56]]     [1, 64, 56, 56]          256      
      ReLU-3         [[1, 256, 56, 56]]    [1, 256, 56, 56]          0       
     Conv2D-7        [[1, 64, 56, 56]]     [1, 64, 56, 56]        36,864     
   BatchNorm2D-7     [[1, 64, 56, 56]]     [1, 64, 56, 56]          256      
     Conv2D-8        [[1, 64, 56, 56]]     [1, 256, 56, 56]       16,384     
   BatchNorm2D-8     [[1, 256, 56, 56]]    [1, 256, 56, 56]        1,024     
 BottleneckBlock-2   [[1, 256, 56, 56]]    [1, 256, 56, 56]          0       
     Conv2D-9        [[1, 256, 56, 56]]    [1, 64, 56, 56]        16,384     
   BatchNorm2D-9     [[1, 64, 56, 56]]     [1, 64, 56, 56]          256      
      ReLU-4         [[1, 256, 56, 56]]    [1, 256, 56, 56]          0       
     Conv2D-10       [[1, 64, 56, 56]]     [1, 64, 56, 56]        36,864     
  BatchNorm2D-10     [[1, 64, 56, 56]]     [1, 64, 56, 56]          256      
     Conv2D-11       [[1, 64, 56, 56]]     [1, 256, 56, 56]       16,384     
  BatchNorm2D-11     [[1, 256, 56, 56]]    [1, 256, 56, 56]        1,024     
 BottleneckBlock-3   [[1, 256, 56, 56]]    [1, 256, 56, 56]          0       
     Conv2D-13       [[1, 256, 56, 56]]    [1, 128, 56, 56]       32,768     
  BatchNorm2D-13     [[1, 128, 56, 56]]    [1, 128, 56, 56]         512      
      ReLU-5         [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-14       [[1, 128, 56, 56]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-14     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-15       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-15     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
     Conv2D-12       [[1, 256, 56, 56]]    [1, 512, 28, 28]       131,072    
  BatchNorm2D-12     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
 BottleneckBlock-4   [[1, 256, 56, 56]]    [1, 512, 28, 28]          0       
     Conv2D-16       [[1, 512, 28, 28]]    [1, 128, 28, 28]       65,536     
  BatchNorm2D-16     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
      ReLU-6         [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-17       [[1, 128, 28, 28]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-17     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-18       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-18     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
 BottleneckBlock-5   [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-19       [[1, 512, 28, 28]]    [1, 128, 28, 28]       65,536     
  BatchNorm2D-19     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
      ReLU-7         [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-20       [[1, 128, 28, 28]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-20     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-21       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-21     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
 BottleneckBlock-6   [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-22       [[1, 512, 28, 28]]    [1, 128, 28, 28]       65,536     
  BatchNorm2D-22     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
      ReLU-8         [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-23       [[1, 128, 28, 28]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-23     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-24       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-24     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
 BottleneckBlock-7   [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-25       [[1, 512, 28, 28]]    [1, 128, 28, 28]       65,536     
  BatchNorm2D-25     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
      ReLU-9         [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-26       [[1, 128, 28, 28]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-26     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-27       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-27     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
 BottleneckBlock-8   [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-28       [[1, 512, 28, 28]]    [1, 128, 28, 28]       65,536     
  BatchNorm2D-28     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
      ReLU-10        [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-29       [[1, 128, 28, 28]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-29     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-30       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-30     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
 BottleneckBlock-9   [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-31       [[1, 512, 28, 28]]    [1, 128, 28, 28]       65,536     
  BatchNorm2D-31     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
      ReLU-11        [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-32       [[1, 128, 28, 28]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-32     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-33       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-33     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
BottleneckBlock-10   [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-34       [[1, 512, 28, 28]]    [1, 128, 28, 28]       65,536     
  BatchNorm2D-34     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
      ReLU-12        [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-35       [[1, 128, 28, 28]]    [1, 128, 28, 28]       147,456    
  BatchNorm2D-35     [[1, 128, 28, 28]]    [1, 128, 28, 28]         512      
     Conv2D-36       [[1, 128, 28, 28]]    [1, 512, 28, 28]       65,536     
  BatchNorm2D-36     [[1, 512, 28, 28]]    [1, 512, 28, 28]        2,048     
BottleneckBlock-11   [[1, 512, 28, 28]]    [1, 512, 28, 28]          0       
     Conv2D-38       [[1, 512, 28, 28]]    [1, 256, 28, 28]       131,072    
  BatchNorm2D-38     [[1, 256, 28, 28]]    [1, 256, 28, 28]        1,024     
      ReLU-13       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-39       [[1, 256, 28, 28]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-39     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-40       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-40    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
     Conv2D-37       [[1, 512, 28, 28]]   [1, 1024, 14, 14]       524,288    
  BatchNorm2D-37    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-12   [[1, 512, 28, 28]]   [1, 1024, 14, 14]          0       
     Conv2D-41      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-41     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-14       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-42       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-42     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-43       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-43    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-13  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-44      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-44     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-15       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-45       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-45     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-46       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-46    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-14  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-47      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-47     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-16       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-48       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-48     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-49       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-49    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-15  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-50      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-50     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-17       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-51       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-51     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-52       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-52    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-16  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-53      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-53     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-18       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-54       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-54     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-55       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-55    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-17  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-56      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-56     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-19       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-57       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-57     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-58       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-58    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-18  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-59      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-59     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-20       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-60       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-60     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-61       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-61    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-19  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-62      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-62     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-21       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-63       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-63     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-64       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-64    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-20  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-65      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-65     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-22       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-66       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-66     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-67       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-67    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-21  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-68      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-68     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-23       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-69       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-69     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-70       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-70    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-22  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-71      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-71     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-24       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-72       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-72     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-73       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-73    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-23  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-74      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-74     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-25       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-75       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-75     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-76       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-76    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-24  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-77      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-77     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-26       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-78       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-78     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-79       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-79    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-25  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-80      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-80     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-27       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-81       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-81     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-82       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-82    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-26  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-83      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-83     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-28       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-84       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-84     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-85       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-85    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-27  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-86      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-86     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-29       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-87       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-87     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-88       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-88    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-28  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-89      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-89     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-30       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-90       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-90     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-91       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-91    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-29  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-92      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-92     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-31       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-93       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-93     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-94       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-94    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-30  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-95      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-95     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-32       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-96       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-96     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
     Conv2D-97       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-97    [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-31  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-98      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-98     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-33       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
     Conv2D-99       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-99     [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-100       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-100   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-32  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-101      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-101    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-34       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-102       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-102    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-103       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-103   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-33  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-104      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-104    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-35       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-105       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-105    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-106       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-106   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-34  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-107      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-107    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-36       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-108       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-108    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-109       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-109   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-35  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-110      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-110    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-37       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-111       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-111    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-112       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-112   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-36  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-113      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-113    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-38       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-114       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-114    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-115       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-115   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-37  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-116      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-116    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-39       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-117       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-117    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-118       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-118   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-38  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-119      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-119    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-40       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-120       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-120    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-121       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-121   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-39  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-122      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-122    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-41       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-123       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-123    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-124       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-124   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-40  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-125      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-125    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-42       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-126       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-126    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-127       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-127   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-41  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-128      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-128    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-43       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-129       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-129    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-130       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-130   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-42  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-131      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-131    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-44       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-132       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-132    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-133       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-133   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-43  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-134      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-134    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-45       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-135       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-135    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-136       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-136   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-44  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-137      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-137    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-46       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-138       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-138    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-139       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-139   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-45  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-140      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-140    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-47       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-141       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-141    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-142       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-142   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-46  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-143      [[1, 1024, 14, 14]]    [1, 256, 14, 14]       262,144    
  BatchNorm2D-143    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
      ReLU-48       [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-144       [[1, 256, 14, 14]]    [1, 256, 14, 14]       589,824    
  BatchNorm2D-144    [[1, 256, 14, 14]]    [1, 256, 14, 14]        1,024     
    Conv2D-145       [[1, 256, 14, 14]]   [1, 1024, 14, 14]       262,144    
  BatchNorm2D-145   [[1, 1024, 14, 14]]   [1, 1024, 14, 14]        4,096     
BottleneckBlock-47  [[1, 1024, 14, 14]]   [1, 1024, 14, 14]          0       
    Conv2D-147      [[1, 1024, 14, 14]]    [1, 512, 14, 14]       524,288    
  BatchNorm2D-147    [[1, 512, 14, 14]]    [1, 512, 14, 14]        2,048     
      ReLU-49        [[1, 2048, 7, 7]]     [1, 2048, 7, 7]           0       
    Conv2D-148       [[1, 512, 14, 14]]     [1, 512, 7, 7]       2,359,296   
  BatchNorm2D-148     [[1, 512, 7, 7]]      [1, 512, 7, 7]         2,048     
    Conv2D-149        [[1, 512, 7, 7]]     [1, 2048, 7, 7]       1,048,576   
  BatchNorm2D-149    [[1, 2048, 7, 7]]     [1, 2048, 7, 7]         8,192     
    Conv2D-146      [[1, 1024, 14, 14]]    [1, 2048, 7, 7]       2,097,152   
  BatchNorm2D-146    [[1, 2048, 7, 7]]     [1, 2048, 7, 7]         8,192     
BottleneckBlock-48  [[1, 1024, 14, 14]]    [1, 2048, 7, 7]           0       
    Conv2D-150       [[1, 2048, 7, 7]]      [1, 512, 7, 7]       1,048,576   
  BatchNorm2D-150     [[1, 512, 7, 7]]      [1, 512, 7, 7]         2,048     
      ReLU-50        [[1, 2048, 7, 7]]     [1, 2048, 7, 7]           0       
    Conv2D-151        [[1, 512, 7, 7]]      [1, 512, 7, 7]       2,359,296   
  BatchNorm2D-151     [[1, 512, 7, 7]]      [1, 512, 7, 7]         2,048     
    Conv2D-152        [[1, 512, 7, 7]]     [1, 2048, 7, 7]       1,048,576   
  BatchNorm2D-152    [[1, 2048, 7, 7]]     [1, 2048, 7, 7]         8,192     
BottleneckBlock-49   [[1, 2048, 7, 7]]     [1, 2048, 7, 7]           0       
    Conv2D-153       [[1, 2048, 7, 7]]      [1, 512, 7, 7]       1,048,576   
  BatchNorm2D-153     [[1, 512, 7, 7]]      [1, 512, 7, 7]         2,048     
      ReLU-51        [[1, 2048, 7, 7]]     [1, 2048, 7, 7]           0       
    Conv2D-154        [[1, 512, 7, 7]]      [1, 512, 7, 7]       2,359,296   
  BatchNorm2D-154     [[1, 512, 7, 7]]      [1, 512, 7, 7]         2,048     
    Conv2D-155        [[1, 512, 7, 7]]     [1, 2048, 7, 7]       1,048,576   
  BatchNorm2D-155    [[1, 2048, 7, 7]]     [1, 2048, 7, 7]         8,192     
BottleneckBlock-50   [[1, 2048, 7, 7]]     [1, 2048, 7, 7]           0       
AdaptiveAvgPool2D-1  [[1, 2048, 7, 7]]     [1, 2048, 1, 1]           0       
     Linear-1           [[1, 2048]]            [1, 173]           354,477    
===============================================================================
Total params: 58,649,709
Trainable params: 58,346,861
Non-trainable params: 302,848
-------------------------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 552.42
Params size (MB): 223.73
Estimated Total Size (MB): 776.72
-------------------------------------------------------------------------------

登录后复制

{'total_params': 58649709, 'trainable_params': 58346861}

登录后复制

（2）定义损失函数和准确率

可图大模型

可图大模型（Kolors）是快手大模型团队自研打造的文生图AI大模型

查看详情

这次使用的是交叉熵损失函数，该函数在分类任务上比较常用。
定义了一个损失函数之后，还有对它求平均值，因为定义的是一个Batch的损失值。
同时我们还可以定义一个准确率函数，这个可以在我们训练的时候输出分类的准确率。

（3）定义优化方法

这次我们使用的是Adam优化方法，同时指定学习率为0.001

Step3.训练模型与训练评估

分为三段式来训练样本：

0-19epoch 学习率0.01

训练集：loss迅速下降，正确率经过二十轮迭代上升至0.80929
测试集：loss迅速下降，正确率迅速上升至0.7384

从0-19的16开始断点续训14个epoch，学习率为0.001

训练集：loss震荡下降，正确率从0.80929上升至0.85714
loss震荡，正确率从0.7384上升至0.7592

选取16+14=30的数据，断点续训11个epoch，学习率为0.0001

训练集：loss震荡下降,正确率从0.85714上升至0.92927
测试集：正确率趋于平稳，正确率取最高值0.82684，即第7个epoch

考虑到训练集持续上升而测试集接近收敛，继续训练会导致过拟合，故训练到此结束。

In [ ]

model.prepare(optimizer=paddle.optimizer.Adam(learning_rate=0.001, parameters=model.parameters()),
              loss=paddle.nn.CrossEntropyLoss(),
              metrics=paddle.metric.Accuracy())# 训练可视化VisualDL工具的回调函数visualdl = paddle.callbacks.VisualDL(log_dir='visualdl_log')# 启动模型全流程训练model.fit(train_dataset,  # 训练数据集
          eval_dataset,  # 评估数据集
          epochs=20,  # 总的训练轮次
          batch_size=64,  # 批次计算的样本量大小
          shuffle=True,  # 是否打乱样本集
          verbose=1,  # 日志展示格式
          save_dir='0_20_resnet152_0.001',  # 分阶段的训练模型存储路径
          callbacks=[visualdl])     # 回调函数使用

登录后复制

The loss value printed in the log is the current step, and the metric is the average value of previous steps.
Epoch 1/20
step 1378/1378 [==============================] - loss: 2.0446 - acc: 0.2861 - 538ms/step         
save checkpoint at /home/aistudio/0_20_resnet152_0.001/0
Eval begin...
step 345/345 [==============================] - loss: 1.2795 - acc: 0.3776 - 256ms/step         
Eval samples: 22049
Epoch 2/20
step 1378/1378 [==============================] - loss: 1.5286 - acc: 0.5133 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/1
Eval begin...
step 345/345 [==============================] - loss: 0.7893 - acc: 0.5104 - 239ms/step         
Eval samples: 22049
Epoch 3/20
step 1378/1378 [==============================] - loss: 1.1590 - acc: 0.5851 - 524ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/2
Eval begin...
step 345/345 [==============================] - loss: 0.6739 - acc: 0.5879 - 240ms/step         
Eval samples: 22049
Epoch 4/20
step 1378/1378 [==============================] - loss: 1.1931 - acc: 0.6275 - 522ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/3
Eval begin...
step 345/345 [==============================] - loss: 0.2853 - acc: 0.6071 - 237ms/step         
Eval samples: 22049
Epoch 5/20
step 1378/1378 [==============================] - loss: 1.3321 - acc: 0.6558 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/4
Eval begin...
step 345/345 [==============================] - loss: 0.5777 - acc: 0.6388 - 242ms/step         
Eval samples: 22049
Epoch 6/20
step 1378/1378 [==============================] - loss: 1.4460 - acc: 0.6820 - 524ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/5
Eval begin...
step 345/345 [==============================] - loss: 0.5820 - acc: 0.6541 - 244ms/step         
Eval samples: 22049
Epoch 7/20
step 1378/1378 [==============================] - loss: 1.0251 - acc: 0.6960 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/6
Eval begin...
step 345/345 [==============================] - loss: 0.3825 - acc: 0.6865 - 239ms/step         
Eval samples: 22049
Epoch 8/20
step 1378/1378 [==============================] - loss: 0.8302 - acc: 0.7145 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/7
Eval begin...
step 345/345 [==============================] - loss: 0.0702 - acc: 0.6644 - 240ms/step         
Eval samples: 22049
Epoch 9/20
step 1378/1378 [==============================] - loss: 1.0031 - acc: 0.7274 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/8
Eval begin...
step 345/345 [==============================] - loss: 0.3258 - acc: 0.6950 - 240ms/step         
Eval samples: 22049
Epoch 10/20
step 1378/1378 [==============================] - loss: 0.9562 - acc: 0.7366 - 522ms/step         
save checkpoint at /home/aistudio/0_20_resnet152_0.001/9
Eval begin...
step 345/345 [==============================] - loss: 0.2758 - acc: 0.7111 - 240ms/step         
Eval samples: 22049
Epoch 11/20
step 1378/1378 [==============================] - loss: 0.9920 - acc: 0.7492 - 524ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/10
Eval begin...
step 345/345 [==============================] - loss: 0.1771 - acc: 0.7077 - 240ms/step         
Eval samples: 22049
Epoch 12/20
step 1378/1378 [==============================] - loss: 0.8259 - acc: 0.7558 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/11
Eval begin...
step 345/345 [==============================] - loss: 0.1928 - acc: 0.7075 - 241ms/step         
Eval samples: 22049
Epoch 13/20
step 1378/1378 [==============================] - loss: 0.8234 - acc: 0.7649 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/12
Eval begin...
step 345/345 [==============================] - loss: 0.4049 - acc: 0.7215 - 243ms/step         
Eval samples: 22049
Epoch 14/20
step 1378/1378 [==============================] - loss: 0.7254 - acc: 0.7730 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/13
Eval begin...
step 345/345 [==============================] - loss: 0.3297 - acc: 0.7173 - 239ms/step         
Eval samples: 22049
Epoch 15/20
step 1378/1378 [==============================] - loss: 0.7475 - acc: 0.7817 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/14
Eval begin...
step 345/345 [==============================] - loss: 0.1445 - acc: 0.7199 - 240ms/step         
Eval samples: 22049
Epoch 16/20
step 1378/1378 [==============================] - loss: 0.7856 - acc: 0.7905 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/15
Eval begin...
step 345/345 [==============================] - loss: 0.3110 - acc: 0.7332 - 238ms/step         
Eval samples: 22049
Epoch 17/20
step 1378/1378 [==============================] - loss: 1.0853 - acc: 0.7945 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/16
Eval begin...
step 345/345 [==============================] - loss: 0.2105 - acc: 0.7461 - 240ms/step         
Eval samples: 22049
Epoch 18/20
step 1378/1378 [==============================] - loss: 1.1084 - acc: 0.8027 - 523ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/17
Eval begin...
step 345/345 [==============================] - loss: 0.1809 - acc: 0.7313 - 241ms/step         
Eval samples: 22049
Epoch 19/20
step 1378/1378 [==============================] - loss: 1.1126 - acc: 0.8079 - 525ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/18
Eval begin...
step 345/345 [==============================] - loss: 0.0287 - acc: 0.7268 - 241ms/step         
Eval samples: 22049
Epoch 20/20
step 1378/1378 [==============================] - loss: 0.6276 - acc: 0.8125 - 524ms/step        
save checkpoint at /home/aistudio/0_20_resnet152_0.001/19
Eval begin...
step 345/345 [==============================] - loss: 0.2583 - acc: 0.7326 - 240ms/step         
Eval samples: 22049
save checkpoint at /home/aistudio/0_20_resnet152_0.001/final

登录后复制

In [ ]

# visualdl --logdir=visualdl_log/ --port=8040# 终端运行此代码

登录后复制

File "/tmp/ipykernel_101/1350602942.py", line 1 visualdl --logdir=visualdl_log/ --port=8080 ^ SyntaxError: can't assign to operator

In [6]

model.save('infer/mnist',training=False)  # 保存模型

登录后复制

---------------------------------------------------------------------------RuntimeError Traceback (most recent call last)/tmp/ipykernel_129/3232998588.py in <module> ----> 1 model.save('infer/mnist',training=False) # 保存模型/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/hapi/model.py in save(self, path, training) 1234 if ParallelEnv().local_rank == 0: 1235 if not training: -> 1236 self._save_inference_model(path) 1237 else:1238 self._adapter.save(path) /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/hapi/model.py in _save_inference_model(self, path) 1981 if self._input_info is None: # No provided or inferred 1982 raise RuntimeError( -> 1983 "Saving inference model needs 'inputs' or running before saving. Please specify 'inputs' in Model initialization or input training data and perform a training for shape derivation." 1984 ) 1985 if self._is_shape_inferred:RuntimeError: Saving inference model needs 'inputs' or running before saving. Please specify 'inputs' in Model initialization or input training data and perform a training for shape derivation.

Step5.模型预测

读取模型参数
对预测图片进行预处理
开始预测

In [ ]

from PIL import Imageimport paddle.vision.transforms as T

model_state_dict = paddle.load('50epoches_chk/final.pdparams')  # 读取模型model = paddle.vision.models.resnet50(num_classes=173)
model.set_state_dict(model_state_dict)
model.eval()

image_file = './splitedDataset/train/108/5_41.jpg'# image_file = './splitedDataset/Yu-Shiang Shredded Pork.webp'# braised pork in brown sauce.jfif \ 1.webp \ rice.jfif \ Yu-Shiang Shredded Pork.webptransforms = T.Compose([
    T.RandomResizedCrop((224, 224)),  # 随机裁剪大小，裁剪地方不同等于间接增加了数据样本 300*300-224*224
    T.ToTensor(),  # 数据的格式转换和标准化 HWC => CHW
    T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])  # 图像归一化])
img = Image.open(image_file)  # 读取图片if img.mode != 'RGB':
    img = img.convert('RGB')
img = transforms(img)
img = paddle.unsqueeze(img, axis=0)

foodLabel = './FoodList.txt'foodList = []with open(foodLabel) as f:    for line in f.readlines():
        info = line.split('\t')        if len(info) > 0:
            foodList.append(info)

ceshi = model(img)  # 测试print('预测的结果为:', np.argmax(ceshi.numpy()), foodList[np.argmax(ceshi.numpy())-1])  # 获取值Image.open(image_file)  # 显示图片

登录后复制

预测的结果为: 108 ['Four-Joy Meatballs\n']

登录后复制

<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=224x224 at 0x7FAC889342D0>

登录后复制

断点续训：

在训练初期，希望训练速度快，使用0.01的学习率。在此基础上需要更小的学习率继续训练，以获得更准确的正确率。

读取模型参数
读取优化器参数
对预测图片进行预处理
开始预测

In [ ]

import paddle.vision.transforms as Timport paddlefrom dataset import FoodDataset

train_data = './train_data1.txt'validation_data = './valida_data1.txt'train_dataset = FoodDataset(train_data)
eval_dataset = FoodDataset(validation_data)

network = paddle.vision.models.resnet101(num_classes=173)
params_dict = paddle.load('15_15_resnet101_0.0001/final.pdparams')
network.set_state_dict(params_dict)
model = paddle.Model(network)

opt = paddle.optimizer.Adam(learning_rate=0.0001, parameters=model.parameters())
opt_dict = paddle.load('15_15_resnet101_0.0001/final.pdopt')
opt.set_state_dict(opt_dict)

model.prepare(optimizer=opt,
              loss=paddle.nn.CrossEntropyLoss(),
              metrics=paddle.metric.Accuracy())# 训练可视化VisualDL工具的回调函数visualdl = paddle.callbacks.VisualDL(log_dir='visualdl_log')# 启动模型全流程训练model.fit(train_dataset,  # 训练数据集
          eval_dataset,  # 评估数据集
          epochs=25,  # 总的训练轮次
          batch_size=64,  # 批次计算的样本量大小
          shuffle=True,  # 是否打乱样本集
          verbose=1,  # 日志展示格式
          save_dir='30_25_resnet101',  # 分阶段的训练模型存储路径
          callbacks=[visualdl])     # 回调函数使用

登录后复制

W0129 23:08:45.468212  6682 device_context.cc:447] Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 10.1, Runtime API Version: 10.1
W0129 23:08:45.472950  6682 device_context.cc:465] device: 0, cuDNN Version: 7.6.

登录后复制

The loss value printed in the log is the current step, and the metric is the average value of previous steps.
Epoch 1/25

登录后复制

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/utils.py:77: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working
  return (isinstance(seq, collections.Sequence) and
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/nn/layer/norm.py:653: UserWarning: When training, we now always track global mean and variance.
  "When training, we now always track global mean and variance.")

登录后复制

step   30/1378 [..............................] - loss: 0.4691 - acc: 0.8672 - ETA: 10:18 - 459ms/st

登录后复制

三、项目总结

实验结果本次实验基于paddlepaddle，利用VireoFood-172 数据集来实现食品识别。网络模型采用的是paddlepaddle自带的ResNet网络模型，可以解决深层网络梯度消失的问题。项目分为三段式来训练样本，第一段为0-19epoch 学习率为0.01，第二段为从0-19的16开始断点续训14个epoch，学习率为0.001，第三段选取16+14=30的数据，断点续训11个epoch，学习率为0.0001，训练结束时测试集正确率达到0.82684，训练集正确率达到0.92927。
实验分析项目一开始选用的是ResNet50模型，但是效果并不理想，因此我们尝试变换模型增加模型深度重新进行训练，选用了ResNet101模型和ResNet152模型，发现随着模型深度加深，训练集和测试集的正确率确实有明显的提升。
后续计划此项目测试集的最高正确率也只能达到0.82684，因此在后续的工作中，尝试使用其他的网络模型，或者继续调节网络模型的超参数，以此来提高训练集和测试集的正确率，这是后续的工作计划。

以上就是基于ResNet-152模型的172种食物图像分类（教育部产学合作协同育人项目）的详细内容，更多请关注php中文网其它相关文章！