3.3 卷积层实现

最新推荐文章于 2024-01-07 08:13:01 发布

liuxiao789

最新推荐文章于 2024-01-07 08:13:01 发布

阅读量598

点赞数

分类专栏： Pytorch框架教程

本文链接：https://blog.csdn.net/liuxiao789/article/details/109061004

版权

Pytorch框架教程专栏收录该内容

15 篇文章 0 订阅

订阅专栏

本文深入探讨了卷积神经网络中的2D卷积和转置卷积操作。通过设置不同参数，展示了它们如何影响图像特征的提取和尺寸变化。使用PyTorch实现并可视化了卷积过程，揭示了卷积层如何从原始RGB图像中生成不同的特征图。同时，文章还介绍了如何将张量转换回PIL图像以便于观察和理解。

摘要由CSDN通过智能技术生成

import os
import torch.nn as nn
from PIL import Image
from torchvision import transforms
from matplotlib import pyplot as plt
from tools.common_tools import transform_invert, set_seed

set_seed(2)  # 设置随机种子---》权值初始化1,2，3
'''
eg: 3个不同的卷积核权值：不同的特征检测器--》出现3个不同的特征图
通常在卷积层中设置多个卷积核，提取不同的特征
'''

# ================================= load img ==================================
# 【1】读取一张RGB图像，再用PIL格式读取进来
path_img = os.path.join(os.path.dirname(os.path.abspath(__file__)), "lena.png")
img = Image.open(path_img).convert('RGB')  # 0~255

# convert to tensor
img_transform = transforms.Compose([transforms.ToTensor()])
img_tensor = img_transform(img)
img_tensor.unsqueeze_(dim=0)    # C*H*W to B*C*H*W

# ================================= 【1】create convolution layer ==================================

# ================ 2d
# flag = 1
flag = 0
if flag:
    conv_layer = nn.Conv2d(3, 1, 3)   # input:(i, o, size) weights:(o, i , h, w)
    nn.init.xavier_normal_(conv_layer.weight.data)
    # 对conv_layer进行初始化

    # calculation-->把图像张量输入到卷积层----》得到输出
    img_conv = conv_layer(img_tensor)

# ================ 【2】transposed
flag = 1
# flag = 0
if flag:
    conv_layer = nn.ConvTranspose2d(3, 1, 3, stride=2)   # input:(i, o, size)
    nn.init.xavier_normal_(conv_layer.weight.data)

    # calculation
    img_conv = conv_layer(img_tensor)


# ================================= visualization ==================================
print("卷积前尺寸:{}\n卷积后尺寸:{}".format(img_tensor.shape, img_conv.shape))

# 对transforms进行逆操作，将张量转换成PIL图像
img_conv = transform_invert(img_conv[0, 0:1, ...], img_transform)
img_raw = transform_invert(img_tensor.squeeze(), img_transform)
plt.subplot(122).imshow(img_conv, cmap='gray')
plt.subplot(121).imshow(img_raw)
plt.show()

'''
卷积前尺寸:torch.Size([1, 3, 512, 512])
卷积后尺寸:torch.Size([1, 1, 510, 510])
'''

    def __init__(self, in_channels, out_channels, kernel_size, stride=1,
                 padding=0, output_padding=0, groups=1, bias=True,
                 dilation=1, padding_mode='zeros'):
        kernel_size = _pair(kernel_size)
        stride = _pair(stride)
        padding = _pair(padding)
        dilation = _pair(dilation)
        output_padding = _pair(output_padding)
        super(ConvTranspose2d, self).__init__(
            in_channels, out_channels, kernel_size, stride, padding, dilation,
            True, output_padding, groups, bias, padding_mode)