This project involves training a deep learning model to drive a car in a simulator using behavioral cloning. The model learns from human driving data and predicts steering angles based on the input images.

The project consists of two main parts:

1. Training the Model: A Convolutional Neural Network (CNN) is trained on images captured by the car's camera along with the corresponding steering angles.
2. Running the Model: The trained model is deployed in a simulator where it controls the car based on the camera feed in real-time.

• behavioural_Cloning.ipynb: Jupyter Notebook containing the model architecture, training process, and data preprocessing steps.
• drive.py: Python script used to interface with the simulator and drive the car using the trained model.
• model.h5: The trained model file (not included, must be generated from the notebook or downloaded).

To run this project, you will need the following dependencies:

• Python 3.7+
• TensorFlow 2.x
• Keras
• Flask
• eventlet
• numpy
• Pillow
• OpenCV
• socketio

The model is trained using a Convolutional Neural Network (CNN) architecture, similar to NVIDIA's model for self-driving cars. The following preprocessing steps are applied to the images:

• Cropping: The image is cropped to remove irrelevant parts, such as the sky and car hood.
• Color Space Conversion: The image is converted to the YUV color space, which is more suitable for computer vision tasks.
• Gaussian Blur: A Gaussian blur is applied to the image to reduce noise and smooth the image.
• Resizing: The image is resized to 200x66 pixels, a size compatible with the input layer of the CNN.
• Normalization: The image pixel values are normalized to a range between 0 and 1, improving the model's convergence during training.

To train the model, run the notebook behavioural_Cloning.ipynb. The notebook contains the following key sections:

1. Data Loading and Augmentation: Load the training data and apply augmentation techniques such as flipping and shifting to increase the dataset's diversity.
2. Model Architecture Definition: Define the CNN architecture, including convolutional layers, dropout layers, and dense layers.
3. Model Training and Validation: Train the model on the preprocessed data and validate its performance using a separate validation set.

Once the model is trained, it can be used to drive the car in the simulator. The drive.py script is responsible for this task. It loads the trained model, preprocesses the incoming images from the simulator, and sends the predicted steering angle back to the simulator.

To drive the car using the trained model, execute the following command in your terminal:

python drive.py model.h5

We will use a driving simulator built in Unity, provided for free by Udacity, that can be found here:

https://github.com/udacity/self-driving-car-sim

after downloading run

./Default\ Linux\ desktop\ Universal.x86_64 

https://www.youtube.com/watch?v=CEgN-XkuUJE