Real-time Object Detection with Audio Feedback (GPU ACCELERATED USING CUDA)

Overview

This Python program performs real-time object detection using a webcam feed, leveraging the YOLOv8 model. It provides audio feedback for detected objects and offers directional guidance to avoid obstacles.

Features

• Real-time object detection using YOLOv8
• Audio feedback for detected objects
• Directional guidance to avoid obstacles
• CUDA support for GPU acceleration

Requirements

• Python 3.6+
• CUDA-capable GPU (optional, for improved performance)

Dependencies

• OpenCV (cv2)
• Ultralytics YOLO
• PyTorch
• pyttsx3

Installation

1. Clone this repository or download the script.

2. Install the required dependencies:

   pip install opencv-python ultralytics torch pyttsx3
   

3. Download the YOLOv8 model weights:

   • The script uses yolov8n.pt by default. You can download it from the Ultralytics YOLO repository.
   • Place the model file in the same directory as the script.

CUDA Setup (for GPU acceleration)

To enable CUDA for GPU acceleration:

1. Ensure you have a CUDA-capable GPU.
2. Install the CUDA Toolkit from the NVIDIA website.
3. Install the cuDNN library from the NVIDIA Developer website.
4. Install the CUDA-enabled version of PyTorch:

   pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
   

   (Replace cu118 with your CUDA version if different)

The script will automatically use CUDA if available.

Want to know more of CUDA installation visit ---My CUDA installation guide

Usage

Run the script using Python:

python object-detection.py

• The program will access your default webcam and start detecting objects in real-time.
• Detected objects will be announced via audio feedback.
• Directional guidance will be provided to avoid obstacles.
• Press 'q' to quit the program.

Customization

• Adjust the confidence_threshold variable to change the detection sensitivity.
• Modify the speech_interval to change how often audio feedback is provided.
• Change the YOLOv8 model by replacing yolov8n.pt with other variants like yolov8s.pt or yolov8m.pt for different performance/accuracy trade-offs.

How It Works

1. The script initializes the YOLO model and the text-to-speech engine.
2. It captures frames from the webcam in real-time.
3. Each frame is processed by the YOLO model for object detection.
4. Detected objects are announced via audio, with a cooldown period between announcements.
5. The program analyzes the position of detected objects and provides directional guidance to avoid obstacles.

Limitations

• Audio feedback may overlap if many objects are detected in quick succession.
• The accuracy of object detection depends on the chosen YOLO model and the confidence_threshold.
• Directional guidance is basic and may not account for complex environments.

Contributing

Feel free to fork this project and submit pull requests with improvements or bug fixes.