DEVELOPER STUDENT CLUBS TIET

Autonomous-Driving

This repository contains the code for steering a self-driving vehicle.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.

Installation and Setup

• Fork the repo and clone it.

git clone https://github.com/developer-student-club-thapar/Autonomous-Driving.git

• Go in the repo and setup virtual environment using python -m virtualenv env or by using anaconda
  conda create --name env
• Then activate the environment using source env/Scripts/activate (For Windows Users using CMD or Powershell env/Scripts/activate) or conda activate env
• Then install the necessary required packages from requirements.txt

pip install -r requirements.txt

• Then run pre-commit install. It will install pre-commit hook for various configurations.

Running the demo

Run the following command :

python autolabeler.py --basedir "all" --datafile "angles.csv" --savefile "./output.csv"

• basedir : Path to folder that dataset folder
• datafile : Path to angles.csv
• savefile : Path to store output.csv

Training the Neural Network

• Make sure your dataset-images are in folder named 'all' at the root directory.
• Make sure your angles.csv and actions.csv are in the root directory named 'angles.csv' and 'actions.csv' respectively.
• The models.py contain architectures of 5-different neural networks named:
  • comma.ai
  • pretrained_vgg16
  • nvidia-dave2
  • 3D-CNN
  • CNN LSTM
• Run the following command :

python steering-model.py --model "model_name from above list"

For Example :

python steering-model.py --model "comma.ai"

NOTE : Use exact name as mentioned above to avoid unnecessary bugs.

Contribution to the project

We follow a systematic Git Workflow -

• Create a fork of this repo.
• Clone your fork of your repo on your pc.
• Add Upstream to your clone
• Every change that you do, it has to be on a branch. Commits on master would directly be closed.
• Make sure that before you create a new branch for new changes,syncing with upstream is neccesary.

Commits

• Write clear meaningful git commit messages (Do read this).
• Make sure your PR's description contains GitHub's special keyword references that automatically close the related issue when the PR is merged. (Check this for more info)
• When you make very very minor changes to a PR of yours (like for example fixing a failing Travis build or some small style corrections or minor changes requested by reviewers) make sure you squash your commits afterward so that you don't have an absurd number of commits for a very small fix. (Learn how to squash at here)

LICENSE

This project is licensed under the MIT License - see the LICENSE.md file for details