涉及的adas技术，用到的深度学习框架

[Aster]

碰撞预警： For collision warning, this system uses CenterNet

信号灯检测： an object detection network with ResNet-18 backbone to achieve a good detection speed and acceptable accuracy

距离估计：perspective transform with calibration is used to estimate the distances from system vehicle to other vehicles ahead （base on ResNet-18）

车道线偏离预警：U-Net and ResNet-18 backbone is used

 

[Aster]

This computer receives two inputs: (i) images from a camera, and (ii) car sensor data such as car speed and turn signal.

[Aster]

Object detection need conversion code to ONNX

[Aster]

(i) forward collision warning with forward vehicles and pedestrians, (ii) lane analysis and lane departure warning, (iii) sign detection for maximum speed limit signs and over-speed warning.

[Aster]

There are six main modules in this system, which are: object detection module, lane detection module, sensing module, warning control module, user interface module and simulation

[Aster]

@Aster In this project, I use MobileNetV2 and ResNet-18 as the backbone of CenterNet.

[Aster]

two main components: a CenterNet based object detection neural network and a ResNet-18 based traffic sign classification network.

[Aster]

@Aster fast-lane restNet只选择56行进行点的检测

[Aster]

车道线检测 fast-lane也用的是ResNet18，关于train------需要原图和二值图以及txt标签文件进行模型训练。