Vision Alert — Real-Time Object Detection with Audio Alerts

A computer vision application that detects objects through your camera and provides spatial audio alerts based on where each object is located in the frame (left, center, or right).

Features

• Real-time object detection using YOLOv8 (80 COCO classes)
• Spatial audio alerts — spoken announcements like "person on the left"
• Configurable cooldown — avoids repeating the same alert too often
• Visual overlay — bounding boxes, labels, and a left/center/right zone guide
• Distance estimation — rough near/mid/far classification based on box size

Quick Start

# 1. Install dependencies
pip install -r requirements.txt

# 2. Run the app
python main.py

Controls

Key	Action
q	Quit
m	Mute / unmute audio alerts
+ / -	Increase / decrease detection confidence threshold

Configuration

Edit config.py or pass CLI flags:

python main.py --confidence 0.5 --cooldown 4 --camera 0

Flag	Default	Description
--camera	0	Camera device index
--confidence	0.45	Minimum detection confidence
--cooldown	3	Seconds before re-announcing the same object+zone
--no-audio	false	Start with audio muted

Project Structure

bet320/
├── main.py            # Entry point — capture loop & display
├── detector.py        # YOLOv8 object detection wrapper
├── audio_alert.py     # Text-to-speech spatial alert engine
├── config.py          # Default configuration & CLI arg parsing
├── requirements.txt
└── README.md

Requirements

• Python 3.9+
• Webcam / camera
• macOS (uses say command for TTS; easily adaptable to other OS via pyttsx3)

How It Works

1. Each frame is divided into three vertical zones: Left (0–33%), Center (33–66%), Right (66–100%).
2. YOLOv8 detects objects and returns bounding boxes.
3. The center-x of each bounding box determines which zone the object is in.
4. The relative height of the bounding box estimates distance (near / mid / far).
5. An audio alert is spoken in a background thread, e.g. "car, center, near".
6. A cooldown timer prevents the same object+zone from being repeated too quickly.