🦖 Dino-NEAT

Introduction

The goal of this project is to create a simple implementation of NeuroEvolution of Augmenting Topologies (NEAT) on the popular chrome dinosaur game.

Dataset

Dataset is just a collection of images of cactus, different dinosaur movement and the terrain.

Installation

Clone the repository using the following command

git clone https://github.com/yuziahaque/Dino_NEAT.git

Install the required dependencies using the following command

pip install -r requirements.txt

Usage

A config.txt file is provided in the repository for customization of the run. Pop size of the config.txt file is the number of dinosaurs per generation

Do make changes to other config variables and see if you can produce any exciting results.

To run the algorithm, Go to the directory folder and run the command

python run main.py

NEAT Brief Explanation

NeuroEvolution of Augmenting Topologies (NEAT) is a genetic algorithm (GA) for the generation of evolving artificial neural networks (a neuroevolution technique)

Configuration Information

Network Parameters

• 'num_hidden' = 0 --> The number of hidden layers in the network

• 'num_inputs' = 2 --> The number of input layers in the network

• 'num_outputs' = 1 --> The number of output layers

• The neural network is activated with two inputs and the output the information:

  • The y-coordinate of the dinosaur (dinosaur.rect.y).
  • The distance between the dinosaur and the obstacle (distance(...)).
  • The output of the neural network (a single value) is used to determine whether the dinosaur should jump or not.