[人工智能]转置卷积与卷积的实现(Pytorch)与可视化(Tensorboard)

代码(code)：

import torch.nn as nn
import cv2
from torch.utils.data import DataLoader
from torchvision import transforms
from torch.utils.tensorboard import SummaryWriter

img = cv2.imread("1.jpg")
tran_tensor = transforms.ToTensor()
img = tran_tensor(img)
print("img.shape: ", img.shape)
print("type(img): ", type(img))

img = img.view(1, 3, 850, 850)

dataloader = DataLoader(dataset=img, batch_size=1)

writer = SummaryWriter("logs")

Conv = nn.Conv2d(in_channels=3, out_channels=3, kernel_size=3, stride=2)
ConvTrans = nn.ConvTranspose2d(in_channels=3, out_channels=3, kernel_size=51)

for data in dataloader:
    img = data
    img = img.reshape(3, 850, 850)
    writer.add_image("input", img, 0)

    img = img.reshape(1, 3, 850, 850)
    output_ConvTrans = ConvTrans(img)
    print("output_ConvTrans.shape: ", output_ConvTrans.shape)
    output_ConvTrans = output_ConvTrans.reshape(3, 900, 900)
    writer.add_image("output_ConvTrans", output_ConvTrans, 1)

    out_Conv = Conv(img)
    print("out_Conv.shape :", out_Conv.shape)
    out_Conv = out_Conv.reshape(3, 424, 424)
    writer.add_image("output_Conv", out_Conv, 2)
writer.close()

运行结果(result)：

可视化(Tensorboard)：

转置卷积的特征图与原图对比

The feature map of the transposed convolution is compared with the original image 

卷积的特征图与原图对比

The feature map of the convolution is compared with the original image