self-driving-car-raspberry

General description

This project is a self driving car based on raspberry pi and tensorflow CNN network designed to drive around user created tracks bounds.

Parts used

1. Raspberry Pi model 4B
2. 7" LCD Screen
3. PiCamera V2
4. Generic USB speaker
5. L298N Two channel motor driver
6. 4 Motors powered by 9V batteries
7. Redmi 18W Fast powerbank 2A 5V
8. bodykit + wheels

Setup

to train your own model run following steps:

1. collect samples: python drive-rc-collect-samples.py use commands:

• w -> go forward
• a -> turn left
• d -> turn right
• s -> go backward
• z -> save sample sequence

2. train model, open jupyter notebook: model_training/ai-acr-train-model.ipynb
3. replace model in script and run: python autodrive.py

Instal packages on raspberry

pip install -r requirements.txt download .whl with tensorflow build matching CPU architecture, then pip install tensorflow-2.5.0-cp37-none-linux_armv7l.whl