Real-Time-Object-Detection-with-OpenCV

This project demonstrates real-time object detection using the YOLOv4 (You Only Look Once) model with OpenCV and a webcam. The model detects objects from the COCO dataset (80 classes) such as person, car, dog, bicycle, etc.

PROJECT STRUCTURE

YOLOv4-Object-Detection/
│── yolov4.cfg          # YOLOv4 configuration file
│── yolov4.weights      # YOLOv4 pre-trained weights
│── coco.names          # COCO dataset class labels (80 objects)
│── detect.py           # Main Python script
│── README.md           # Documentation

Dependencies

Python Libraries Install required libraries with:

pip install opencv-python numpy

External Files

Download the following and place them in your project folder:

yolov4.weights (248 MB) yolov4.cfg coco.names

How to Run

1. Clone or download this project folder.
2. Make sure the following files are in the same directory: yolov4.cfg yolov4.weights coco.names detect.py
3. Run the script:

python detect.py

4. The webcam will open, and YOLOv4 will start detecting objects in real-time.
5. Press q to exit the window.

EXAMPLE OUTPUT

Objects detected will be shown with bounding boxes and class labels:

person 0.95
dog 0.88
car 0.79

Notes

Default confidence threshold is 0.5 (50%). Non-Max Suppression (NMS) is applied to avoid overlapping boxes. Works with any webcam accessible to OpenCV.

Future Enhancements

Add GPU (CUDA) support for faster inference. Save detection results to a file. Extend to process videos instead of webcam.