🧠 Self-Driving Car using Udacity Simulator

This project is a deep learning-based self-driving car implementation using the Udacity Self-Driving Car Simulator. A convolutional neural network (CNN) is trained with images from the simulator to predict steering angles for autonomous driving.

🚗 Project Overview

The car receives images from a front-facing camera in the simulator and uses a trained model to predict the steering angle in real time. The model is trained using TensorFlow and follows an end-to-end learning approach inspired by NVIDIA’s self-driving research.

🔧 Technologies Used

• Python 3
• TensorFlow 2.x / Keras
• OpenCV
• NumPy
• Udacity Self-Driving Car Simulator

📦 Installation

1. Clone this repository

git clone [https://github.com/GeneralSubhra/Self-driving-car]
cd self-driving-car-udacity

2. Create and activate a virtual environment (optional but recommended)

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Download the Udacity Simulator

Simulator download: Udacity Self-Driving Car Simulator

🚀 How to Run

📸 Collect Training Data (Optional)

Use the simulator in training mode to drive manually and collect images + steering angles. The data will be saved into a CSV log and an image folder.

🧠 Train the Model

python train.py

This will train the model and save it as model.h5.

🕹️ Drive Using the Trained Model

Run the simulator in autonomous mode, then:

python drive.py model.h5

📁 Project Structure

.
├── data/                   # Collected driving data (images + driving_log.csv)
├── model/                  # Saved model(s)
├── script.py                # Script to drive car in simulator using trained model
├── app.py                # CNN architecture (TensorFlow/Keras)
├── requirements.txt        # List of dependencies
└── README.md               # This file

✅ To-Do / Future Work

• Add lane detection and object avoidance
• Train on multiple tracks
• Implement PID control for smoother driving

🤝 Contributing

Contributions and suggestions are welcome! Please fork the repo and open a pull request.

📄 License

This project is licensed under the MIT License.