README.md

Self-Driving-Car-Neutral-Network

Demo

View here

Detect traffic signs	Run
Obstacle

Table of Contents

1. About The Project
2. Built With
   • Framework and library
   • Hardware
3. Usage
   • Step1 DataCollection
   • Step2 LabelData
   • Step3 Training
4. Reference documents

About The Project

This project uses the deep learning convolutional neural network U-Net-LSTM to help cars recognize their lane along with YOLOv8 to identify signs and vehicle owners.By using U-Net-LSTM the car can identify and draw lanes accurately in up to 96% of labeled images. Something that image processing is almost impossible to do, especially with images with many curves. With this method, vehicles can even drive on lanes without lane markings as long as they are clearly labeled.

Built With

Framework and library

• Pytorch
• Python
• Opencv
• Pyfirmata
• Socket

Hardware

Raspberry pi zero 2 w	Camera 5MP v1.3cho Raspberry Pi
Other

(back to top)

Usage

Step1 DataCollection

• Run main.py to collect lane data as seen by the vehicle through the camera. Below is a link to the dataset used in the project
• Lane data: Link

Step2 LabelData

• Run main.py to label the newly collected data
• Watch the video to label data and training an object recognition model using yolov8
• My traffic sign dataset: Link

Step3 Training

• Run train.py with your new dataset
• You can use my dataset by downloading here and replace folder with the same name

Step4 Predict

• Run predict.py to see how the model can recognize the lane
• Run server.py in your laptop
• Run client.py in CodeInCar dicrectory

(back to top)

Reference documents

Sincere thanks to the team who wrote this paper . This research has been very helpful in creating this project

(back to top)