XAI-in-Self-Driving-Car

This project implements Explainable AI in self driving cars by the use of post-hoc explanations.

---

This repository contains Python scripts for training a machine learning model, running simulations, performing explanations using LIME, and conducting object detection.

Contents

1. model_training.py: Python script for training the machine learning model.
2. drive.py: Python script to run the trained model in a simulation environment.
3. limepro.py: Python script demonstrating LIME (Local Interpretable Model-agnostic Explanations) for explaining model predictions
4. object_detection.py: Python script for performing object detection tasks.

Usage

• Model Training: Use model_training.py to train your machine learning model. Adjust parameters and datasets as needed within the script.

• Simulation: Run simulations using drive.py to evaluate the trained model in a simulated environment such as Udacity.

• Explanations: Explore explanations for model predictions with limepro.py using the LIME framework.

• Object Detection: Perform object detection tasks using object_detection.py.

• Data Collection: The data used to train the model are image files obtained by using the record feature available in both airsim and udacity simulators. The data has to preprocessed and augmented.

Additional Resources

• Udacity Nanodegree Program: The simulation environment used is part of the Udacity Nanodegree program. For more details and to download visit Udacity Nanodegree Program and to download-

  Linux

  Mac

  Windows.

• AirSim Simulator: For simulation purposes, download the AirSim simulator from AirSim GitHub Repository.

---

Both simulation environments are open-source.