一个简单的残差网络,残差部分采用两层卷积结构，任务是是有监督降噪，环境为torch

import torch import glob import scipy import numpy as np import os from scipy.io import loadmat from ResNet import ResidualNet from torch import optim from torchvision.datasets import CIFAR10 from torchvision.transforms import transforms from torch.utils.data import DataLoader from data_generator import DenoisingDataset batch_size = 16 clean_dir = 'D:\\DAS数据集\\duoleixing\\chunjing\\' noise_dir = 'D:\\DAS数据集\\duoleixing\\suijinoise\\' save_dir = 'D:\\DAS数据集\\duoleixing\\suiji_model\\' # 创建网络实例 model = ResidualNet() model.train() criterion = torch.nn.MSELoss() optimizer = optim.Adam(model.parameters(), lr=0.001) #####数据加载函数 file_list = glob.glob(clean_dir + '/*.mat') # get name list of all .png files data1 = [] for i in range(len(file_list)): mat_content = loadmat(file_list[i]) patches = mat_content['chunjing'] data1.append(patches) data1 = np.array(data1, dtype='float32') data1 = np.expand_dims(data1, axis=3) #给每个数据加一个中括号 discard_n = len(data1) - len(data1) // batch_size * batch_size # because of batch namalization data1 = np.delete(data1, range(discard_n), axis=0) print('^_^-clean data finished-^_^') file_list2 = glob.glob(noise_dir + '/*.mat') # get name list of all .png files data2 = [] for i in range(len(file_list2)): mat_content2 = loadmat(file_list2[i]) patches2 = mat_content2['noise'] data2.append(patches2) data2 = np.array(data2, dtype='float32') data2 = np.expand_dims(data2, axis=3) #给每个数据加一个中括号 discard_n = len(data2) - len(data2) // batch_size * batch_size # because of batch namalization data2 = np.delete(data2, range(discard_n), axis=0) print('^_^-noise data finished-^_^') #####数据加载结束 xs = torch.from_numpy(data1.transpose((0, 3, 1, 2))) sigma = torch.from_numpy(data2.transpose((0, 3, 1, 2))) DDataset = DenoisingDataset(xs, sigma) DLoader = DataLoader(dataset=DDataset, num_workers=0, drop_last=True, batch_size=batch_size, shuffle=True) epoch_loss = 0 num_epochs = 100 for epoch in range(num_epochs): for n_count, batch_yx in enumerate(DLoader): optimizer.zero_grad() targets, inputs = batch_yx[1], batch_yx[0] outputs = model(inputs) ##得到输入输出 loss = criterion(outputs, targets) epoch_loss += loss.item() loss.backward() optimizer.step() print(f'Epoch {epoch + 1}/{num_epochs}, Loss: {loss.item()}') torch.save(model, os.path.join(save_dir, 'model_%03d.pth' % (epoch + 1)))