Autonomous Driving Car

End-to-end self-driving car project for the Udacity simulator.

This project first trains a steering model from driving data, then uses that trained model for real-time autonomous driving with path visualization.

Demo

Full demo video: Self Driving Car.mp4

Training

Model training workflow is included in Autonomus_Self_Driving_Car.ipynb, including:

• data loading from driving logs
• steering distribution balancing
• image augmentation and preprocessing
• NVIDIA-style CNN training
• model export

The trained model used for driving is stored at model/model.h5.

Inference and Control

Runtime driving is handled by drive.py:

• receives simulator camera frames
• preprocesses frames and predicts steering
• computes throttle with speed control logic
• renders a live predicted path overlay

Bezier curve utilities for path smoothing are in bezier_path.py.

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Start the Udacity simulator.

Run

python drive.py