Image Processing for Self-Driving in GTA V

This project implements computer vision algorithms to enable basic self-driving capabilities within the GTA V environment. It utilizes real-time image processing to detect lanes and surrounding objects, providing a foundation for automated navigation and obstacle avoidance.

Project Overview

The core objective is to demonstrate how an automated vehicle perceives its environment. By scanning video frames in real-time, the application identifies driving lanes and classifies objects such as cars, pedestrians, and traffic lights. It also includes a proximity warning system to alert when obstacles are too close.

Features

• Real-time Lane Detection: Uses the Hough Transform algorithm to identify and highlight road lanes.
• Object Detection: Implements the YOLOv3 (You Only Look Once) algorithm to detect and classify multiple objects in the frame.
• Proximity Warning: Triggers a "Warning !!!" message on-screen when a detected vehicle (car or truck) is dangerously close.
• GTA V Integration: Designed to capture and process gameplay from GTA V, providing a realistic simulation environment.

Architecture & Algorithms

1. Lane Detection (Hough Transform)

• Canny Edge Detection: Identifies sharp changes in intensity to find potential edges.
• Gaussian Blur: Reduces noise and smoothens the image before edge detection.
• Region of Interest (ROI): Focuses processing on the area of the frame where lanes are most likely to appear.
• Hough Line Transform: Extracts straight lines from the edge-detected image to represent lane boundaries.

2. Object Detection (YOLOv3)

• Deep Learning Model: Utilizes a pre-trained YOLOv3 network for efficient, real-time object detection.
• Bounding Boxes: Draws rectangles around detected objects with classification labels and confidence scores.
• Non-Maximum Suppression (NMS): Filters redundant overlapping boxes to ensure each object is detected only once.

Project Structure

.
├── assets/
│   ├── clips/          # Sample video clips for testing
│   └── coco.names      # COCO dataset class names for YOLO
├── src/
│   ├── lane_detection.py    # Independent lane detection implementation
│   ├── object_detection.py  # Independent object detection implementation
│   └── self_driving.py      # Integrated lane and object detection
├── requirements.txt    # Python dependencies
└── README.md           # Project documentation

Setup & Installation

Prerequisites

• Python 3.x
• GTA V (optional, for real-time testing)

Installation

1. Clone the repository:

   git clone <repository-url>
   cd <repository-directory>

2. Install dependencies:

   pip install -r requirements.txt

3. Download YOLOv3 Weights & Config: The YOLOv3 model files are required for object detection. Download them and place them in the assets/ directory:

   • yolov3.weights
   • yolov3.cfg

Usage

Running with Sample Clips

If you don't have the game, you can run the detection on sample clips provided in assets/clips/:

python src/object_detection.py

Running with GTA V

1. Launch GTA V and set the resolution to 800x600.
2. Position the game window in the top-left corner of your screen.
3. Run the integrated self-driving script:

   python src/self_driving.py

Authors

• Dhruv Pathak
• Vrushit Patel

License

This project is open-source and intended for educational purposes.