autonomous-driving

This is the autonomous driving project of FH Hof. The code here is the result of different thesis of students.

This is the autonomous-driving test car

We tested it using Windows 10 operating system. If you clone it to “D:\Projekte” you will have the least amount of hassle. It should however run on Linux as well.

We have some sub projects with different tasks:

Sub projects

working

If a subprojekt creates files, all of them are stored in this folder. If you clone it it should be empty

pylon-images

Here we manage pictures of pylons and backgrounds.

image-generator

We use the images from pylon images to create training images.

transfer-learning

We build up a neural network using the technique of transfer learning to recognize pylons from the training images. The finished neural network can then be exported as tflite or coral tflite.

trained-networks

The fully trained neural networks are stored here.

car-controller

The code for the raspberry pi and the Arduino can be found here.

video

A video of a test course.

Python Environments

We have a separate anaconda environment for each sub project. The name of the environment is the project name. image-generator is the example environment in the following commands.

To use the commands navigate in a sub project with the anaconda prompt.

create environment

conda create -n transfer-learning python=3.5

list of environments

conda info --envs

install environment

conda env create -f environment.yml

change environment

conda activate image-generator

remove environment

conda remove --name image-generator --all

update environment config

After a change of the environment the changes have to be saved in environment.yml.

conda env export  > environment.yml