毕设&课程作业_基于纯视觉端到端深度学习的自动驾驶系统.zip

# 基于纯视觉端到端深度学习的自动驾驶系统 ## 效果 [](https://youtu.be/7dTMGus53y8) <p align="center"> <img src="https://raw.githubusercontent.com/lyzMaster/DeepGTA5-V2/master/image/1.PNG"> </p> <p align="center"> <img src="https://raw.githubusercontent.com/lyzMaster/DeepGTA5-V2/master/image/2.PNG"> </p> <p align="center"> <img src="https://raw.githubusercontent.com/lyzMaster/DeepGTA5-V2/master/image/3.PNG"> </p> <p align="center"> <img src="https://raw.githubusercontent.com/lyzMaster/DeepGTA5-V2/master/image/4.PNG"> </p> <p align="center"> <img src="https://github.com/lyzMaster/deepgta5/raw/master/models/image/drive.gif"> </p> ## 使用 `python drive.py` 打开[远程遥控器](https://www.pixeldesert.com/console/)，控制开始或结束 #### 总架构 <img src="https://github.com/lyzMaster/deepgta5/raw/master/models/image/full.png"> #### 主模型(NVDIA) <img src="https://github.com/lyzMaster/deepgta5/raw/master/models/image/main_model.png"> #### 路径模型 <img src="https://github.com/lyzMaster/deepgta5/raw/master/models/image/ladar_model.png"> ## 使用的工具 控制GTA5中的汽车运行：使用[DeepGTAV](https://github.com/aitorzip/DeepGTAV)控制车辆速度，以及转弯角度。 绘制卷积神经网络图像：[draw_convnet](https://github.com/gwding/draw_convnet)，keras的plot_model函数 ## 技术详情请查看：项目设计说明书.pdf