♟️ Chess Using ML

A web-based chess application that uses Machine Learning to predict and play chess moves. Built with Streamlit and powered by a trained neural network model.

---

🚀 Features

• Interactive Chess Board: Visual chess board with piece movement
• ML-Powered Opponent: ML opponent that uses a trained neural network to make moves
• Automatic Play: ML automatically responds after each user move
• Multiple Input Methods:
  • Click on move buttons to select your move
  • Enter moves using UCI notation (e.g., e2e4)
• Game Controls: Undo moves, reset game, view move history
• Real-time Game Status: Shows current turn, check/checkmate status
• Move History: Track all moves made during the game

---

🛠️ Technology Stack

• Frontend: Streamlit
• Chess Logic: python-chess library
• ML Framework: TensorFlow/Keras
• Board Encoding: One-hot encoding for neural network input
• Visualization: SVG-based chess board rendering

---

📋 Requirements

streamlit  
chess  
numpy  
tensorflow  
keras  
pandas  
matplotlib

Install all dependencies using:

pip install -r requirements.txt

---

🚀 Installation & Setup

Clone the repository:

git clone https://github.com/kavanatn/Chess-using-ML.git
cd Chess-using-ML

Ensure you have the trained model:

• Make sure chess_model.h5 is in the project directory
• This file contains the trained neural network weights
• Or train it yourself using the provided Google Colab notebook (see below)

Run the application:

streamlit run app.py

Open your browser and go to:

http://localhost:8501

Start playing chess using machine learning!

---

🎮 How to Play

• You play as White (first move)
• Make your move by either:
  • Clicking on the move buttons displayed
  • Entering UCI notation in the text input (e.g., "e2e4")
• ML responds automatically as Black after your move
• Continue playing until checkmate, stalemate, or draw

---

🧠 ML Model Details

The chess engine uses a neural network that:

• Input: One-hot encoded chess board positions (8x8x13 tensor)
• Output: Move evaluation scores
• Training: Trained on chess game data to evaluate board positions
• Move Selection: Chooses moves based on position evaluation

Board Encoding:

• Each square encoded as a 13-dimensional one-hot vector
• Represents 12 piece types (6 white + 6 black) + empty square
• Board converted to 8x8x13 tensor for neural network input

---

📓 Model Training

Model training is done using Google Colab. The notebook is available here:

• train_and_play.ipynb
• Requires data files located in the data/ folder:
  • data/train.csv
  • data/test.csv
• After training, chess_model.h5 will be generated and used by the Streamlit app

To run the notebook in Google Colab:

---

📁 Project Structure

chess-using-ml/
├── app.py                # Main Streamlit application  
├── chess_utils.py        # Chess utility functions and ML model interface  
├── chess_model.h5        # Trained neural network model  
├── requirements.txt      # Python dependencies  
├── train_and_play.ipynb  # Google Colab notebook (training)  
├── data/                 # Training and testing data  
│   ├── train.csv  
│   └── test.csv  
└── README.md             # Project documentation  

---

🎯 Key Functions

• encode_board(): Converts chess board to neural network input format
• one_hot_encode_piece(): Encodes individual pieces as one-hot vectors
• play_nn(): Uses ML model to select best move for given position
• make_ai_move(): Handles automatic ML move execution