Carla使用神经网络训练自动驾驶车辆模型搭建及训练

上一节已经搭建好了一个carla仿真环境&＃xff0c;并且进行了数据采集

现在需要使用采集到的摄像头图片和转角数据进行模型训练

创建dataset类

import numpy as np
import config
from torch.utils.data import Datasetclass TrainDataSet(Dataset):def __init__(self):self.x&＃61;np.load(config.DATASET_PATH&＃43;"x_train.npy")[:config.TRAIN_DATASET_SIZES]self.y&＃61;np.load(config.DATASET_PATH&＃43;"y_train.npy")[:config.TRAIN_DATASET_SIZES]def __len__(self):return len(self.x)def __getitem__(self,index):return (self.x[index],self.y[index])class TestDataSet(Dataset):def __init__(self):self.x &＃61; np.load(config.DATASET_PATH&＃43;&＃39;x_train.npy&＃39;)[config.TRAIN_DATASET_SIZES:]self.y &＃61; np.load(config.DATASET_PATH&＃43;&＃39;y_train.npy&＃39;)[config.TRAIN_DATASET_SIZES:]def __len__(self):return len(self.x)def __getitem__(self, index):return (self.x[index], self.y[index])


模型部分

import torch
import torch.nn as nn
from torchsummary import summaryclass Block(nn.Module):def __init__(self, in_channels, out_channels, maxpool&＃61;False):super().__init__()self.conv &＃61; nn.Conv2d(in_channels, out_channels, 3, 1, 1)self.relu &＃61; nn.ReLU()self.maxpool &＃61; maxpoolself.max_pool &＃61; nn.MaxPool2d(2, 2)def forward(self, x):x &＃61; self.conv(x)x &＃61; self.relu(x)if self.maxpool:x &＃61; self.max_pool(x)return xclass Model(nn.Module):def __init__(self, in_channels&＃61;3, out_size&＃61;1):super().__init__()self.conv_block &＃61; nn.Sequential(Block(in_channels, 16),Block(16, 16, True),Block(16, 32),Block(32, 32, True),Block(32, 64),Block(64, 64, True),Block(64, 128),Block(128, 128, True),Block(128, 128),Block(128, 128, True),)self.fc_block &＃61; nn.Sequential(nn.Linear(1536, 100),nn.ReLU(),nn.Linear(100, 20),nn.ReLU(),nn.Linear(20, out_size))def forward(self, x):x &＃61; self.conv_block(x)x &＃61; x.view(x.size(0), -1)x &＃61; self.fc_block(x)return xif __name__ &＃61;&＃61; &＃39;__main__&＃39;:model &＃61; Model()summary(model, input_size&＃61;(3, 66, 200), device&＃61;&＃39;cpu&＃39;)data &＃61; torch.ones(1, 3, 66, 200)out &＃61; model(data)print(out.shape)


训练部分&＃xff1a;

import torch
import torch.nn as nn
import keyboard
import matplotlib.pyplot as plt
import config
from torch.utils.data import DataLoader
from dataset import TrainDataSet
from model import Modelclass Trainer:def __init__(self, model, train_dataset, model_state&＃61;None):self.model &＃61; modelself.model_state &＃61; model_stateself.train_dataset &＃61; train_datasetself.loss_list &＃61; []self.main()def main(self):torch.manual_seed(config.SEED)torch.cuda.manual_seed(config.SEED)model &＃61; self.modelloss_list &＃61; self.loss_listmodel.to(config.DEVICE)if self.model_state:state &＃61; torch.load(self.model_state)model.load_state_dict(state)model.train()dataloader &＃61; DataLoader(self.train_dataset, batch_size&＃61;config.BATCH_SIZE, shuffle&＃61;True)optimizer &＃61; torch.optim.Adam(model.parameters(), lr&＃61;config.LEARNING_RATE)criterion &＃61; nn.MSELoss()for epoch in range(config.NUM_EPOCH):for x, y in dataloader:x, y &＃61; x.to(config.DEVICE).float(), y.to(config.DEVICE).float()x &＃61; x.reshape(x.size(0), 3, 66, 200)optimizer.zero_grad()out &＃61; model(x)loss &＃61; criterion(out, y)loss.backward()optimizer.step()loss_list.append(loss.item())self._show_loss(loss_list, &＃39;3&＃39;)print(&＃39;epoc[%i/%i] loss&＃61;%.5f&＃39; % (epoch, config.NUM_EPOCH, loss.item()))torch.save(model.state_dict(), config.MODEL_PATH&＃43;&＃39;model_state.pth&＃39;)def _show_loss(self, loss_list, key&＃61;&＃39;3&＃39;):if keyboard.is_pressed(key):plt.plot(loss_list)plt.ylim(0, 0.1)plt.show()if __name__ &＃61;&＃61; &＃39;__main__&＃39;:# Trainer(Model(), TrainDataSet(), model_state&＃61;&＃39;model_state.pth&＃39;)Trainer(Model(), TrainDataSet())


这样就会得到神经网络的权重model_state.pth&＃xff0c;下一节使用训练好的神经网络在carla环境中进行自动驾驶测试