把vgg16用在单通道灰度图上

1.把vgg16用在单通道灰度图上，具体做法就是直接将第一个卷积层的输入通道改为1

附完整代码：

import torch
from torch import nn
from torchvision.models.vgg import vgg16
from PIL import Image
import torchvision.transforms as transforms
img_to_tensor = transforms.ToTensor()
from torch.autograd import Variable
import numpy as np

vgg = vgg16(pretrained=True)
#更改vgg网络的input为单通道
vgg.features[0]=nn.Conv2d(1, 64, kernel_size=3, padding=1)

def inference(model, imgpath):
    model.eval()  # 必需，否则预测结果是错误的

    img = Image.open(imgpath)
    img = img.resize((224, 224))
    tensor = img_to_tensor(img)
    tensor = tensor.resize_(1, 1, 224, 224)

    result = model(Variable(tensor))
    result_npy = result.data.cpu().numpy()  # 将结果传到CPU，并转换为numpy格式
    max_index = np.argmax(result_npy[0])

    return max_index

imgpath = 'F:\pycharmProject\SRGAN-master\grey.png'
print(inference(vgg, imgpath))

2.vgg16模型减半，以pre_trained vgg16权重作为初始权重，重新训练vgg16（思路是这样，我还没运行过，因为还没下载ImageNet数据集，大家借鉴思路就行了）

附完整代码：

import torch
from torch import nn
from torchvision.models.vgg import vgg16
from PIL import Image
import torchvision.transforms as transforms
import torchvision.datasets as dsets
img_to_tensor = transforms.ToTensor()
from torch.autograd import Variable
import numpy as np
import torch.optim as optim

EPOCH=50
BATCH=4
LEARNING_RATE = 0.01

#process data
transform = transforms.Compose([
    transforms.RandomSizedCrop(224),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize(mean = [ 0.485, 0.456, 0.406 ],
                         std  = [ 0.229, 0.224, 0.225 ]),