Computer Vision — YOLOv8 Object Detection (User Manual)

This project contains three simple Python programs that run object detection with YOLOv8 using OpenCV:

• simple_object_detection.py: Detect objects in a single image and show an annotated window.
• object_detection_from_video.py: Detect objects from a video file (e.g., traffic.mp4).
• live_camera_feed.py: Detect objects from your live webcam feed.

A small YOLOv8 model file yolov8n.pt is included for quick tests.

Prerequisites

• Python 3.8–3.12 on Windows
• A working webcam (for live_camera_feed.py)
• A video file (e.g., traffic.mp4) for object_detection_from_video.py

1) Install dependencies

pip install -r requirements.txt

This installs ultralytics (YOLOv8) and opencv-python.

2) Renaming and updating the environment file

Rename the environment files as ".env" from ".env.example" and enter your API_KEYS

Files

• simple_object_detection.py — image inference script
• object_detection_from_video.py — video file inference script
• live_camera_feed.py — webcam/live camera inference script
• yolov8n.pt — YOLOv8n model weights
• traffic.mp4 — example input video
• person-image.jpg, fruits.jpg — example images

How to Run

Make sure your virtual environment is activated if you created one.

1) Single image

Run:

python simple_object_detection.py

Defaults to person-image.jpg. To change the image, edit the script and set the file you want:

image = cv2.imread("person-image.jpg")
# image = cv2.imread("fruits.jpg")

The program will open a window titled "Annotated Image". Press q or close the window to exit.

2) Video file

Run:

python object_detection_from_video.py

• The script reads traffic.mp4 by default:

  cap = cv2.VideoCapture("traffic.mp4")

  Change the path to your own video file if needed.
• The annotated video is resized to 800×600 for display and shown in a window titled "Video Object Detection".
• Press q to quit.

3) Live camera

Run:

python live_camera_feed.py

• Uses the default camera index 0:

  cap = cv2.VideoCapture(0)

  If you have multiple cameras or the default doesn’t work, try 1, 2, etc.
• A window titled "Live Camera Feed" will appear. Press q to quit.

Switching models

yolov8n.pt is the small, fast model. You can switch to other YOLOv8 models (e.g., yolov8s.pt, yolov8m.pt) by placing the weights in the project folder and changing:

model = YOLO("yolov8n.pt")

Tips and Notes

• GPU (CUDA) is used automatically if available via PyTorch; otherwise, inference runs on CPU.
• Larger models may improve accuracy but will be slower.
• For best webcam performance, close other apps that use the camera.

Troubleshooting

• Camera window is black or doesn’t open:
  • Another app may be using the camera. Close it and try again.
  • Try a different camera index: cv2.VideoCapture(1).
  • Ensure Windows camera privacy permissions allow desktop apps.
• Video file won’t open:
  • Check the file path and file name in the script.
  • Try an absolute path like C:\\path\\to\\your_video.mp4.
• Module not found (e.g., cv2):
  • Reinstall dependencies: pip install -r requirements.txt (ensure your venv is active).
• Very slow on CPU:
  • Stick with yolov8n.pt (already the smallest model).
  • Close other CPU-intensive applications.

Exit controls

• All programs use the same exit key: press q while the window is focused.

License

This sample is intended for educational and demonstration purposes.