gritchou / text2play

AI methods showcase

Text2Play 🎮🎨

Bringing Art to Life through AI

Overview

Text2Play is an innovative web application that transforms user prompts into a 2D platform video game. By leveraging AI techniques, it converts text into famous paintings and applies style transfer to create a unique game background.

Features

• Word Vectorization: Utilizes BERT for word embeddings to understand and interpret user prompts.
• Style Transfer: Applies VGG19 with PyTorch to transfer the style of famous paintings onto the game's background.
• Game Generation: Integrates the generated background into a 2D platform game using Phaser.io.

Tech Stack

• Backend: Python, Flask
• AI Models: BERT, VGG19, PyTorch
• Frontend: React, Phaser.io
• Deployment: Google Cloud Run, Docker, MLflow for model orchestration

Installation

1. Clone the repository:

   git clone https://github.com/gritchou/Text2Play.git

2. Navigate to the project directory:

   cd Text2Play

3. Install dependencies:

   make install

4. Preprocess the dataset:

   make preprocess-dataset

5. Run the application:

   make run

Frontend Setup

1. Clone the frontend repository:

   git clone https://github.com/gritchou/Text2Play-frontend.git

2. Navigate to the frontend project directory:

   cd Text2Play-frontend

3. Install frontend dependencies:

   yarn install

4. Start the frontend application:

   yarn start

Usage

1. Enter a prompt describing the scene.
2. The backend processes the prompt using BERT for word vectorization.
3. VGG19 applies style transfer to create a game background.
4. The frontend displays the 2D platform game with the generated background.

Contributors

• Jean-François Grand - LinkedIn | GitHub
• Guillaume Padan - GitHub
• Shuja Khan - GitHub

References

• Neural-Style algorithm
• Neural Transfer Using PyTorch
• BERT Model
• VGG19

License

This project is licensed under the MIT License.