Real-Time Object Detection

Overview

A real-time object detection system built using Python, YOLOv8, and OpenCV.

This project performs real-time detection of the following object classes from a live webcam or video stream: Person, Dog, Cat, and Car. It detects and labels each object, draws bounding boxes, displays confidence scores, counts detected objects, and shows the FPS for performance monitoring.

Tech Stack

• Python
• YOLOv8
• OpenCV

Features

• Real-time object detection
• Detects multiple object types
• Draws bounding boxes around detected objects
• Displays object labels and confidence scores
• Counts the number of detected objects
• Supports live webcam or RTSP video stream
• Displays FPS for performance monitoring

How It Works

1. The camera or video stream is opened using OpenCV.
2. Each video frame is passed to the YOLOv8 model.
3. The model detects objects in the frame.
4. Detected objects are labeled with their class name and confidence score.
5. Bounding boxes are drawn around the detected objects.
6. The total number of detected objects is displayed.
7. The processed frame is shown in real time.

Requirements

• Python 3.x
• Webcam or RTSP stream

Setup

pip install ultralytics opencv-python

Run

python main.py
# or
python3 main.py

Use python or python3 depending on your system.

Notes

• Make sure a webcam or RTSP source is available.
• Press ESC to exit the application.
• The detection accuracy depends on the YOLOv8 model and video quality.

Future Improvements

• Add a graphical user interface
• Save detection results to a file
• Add support for uploading video files
• Improve object counting by category

Challenges

• Implemented multi-threaded frame capture to reduce latency.
• Managed frame buffering to avoid processing delays.
• Optimized the application to run on both CPU and CUDA-supported devices.
• Filtered detections to focus only on the required object classes.