AI USB Camera: Real-Time Object Detection using the YOLOv8 Model in Ubuntu

AI Real-Time Object Detection in Ubuntu Using YOLOv8

AI USB Camera:Install Required Packages in Ubuntu

• Setting up your AI USB camera in the Ubuntu environment.
• Integrating and utilizing the modern YOLOv8 model for real-time object detection.
• Creating your very own Python script that connects everything together.

Object Detection Using USB Camera in Ubuntu – YOLOv8 Recognizes Objects in Real Time

1. Installing the required packages

First, it is necessary to install the required libraries on your system. Open the terminal and enter the following commands:

sudo apt update
sudo apt install python3-pip
pip install ultralytics opencv-python

Update

Install python3

Install ultralytics opencv

If the installation fails, you can try to force the installation with a potential risk of affecting system packages (use with caution):

pip install ultralytics opencv-python --break-system-packages

Install ultralytics opencv --break-system-packages !

Then update the pip package manager:

pip install --upgrade pip

or

pip install --upgrade pip --break-system-packages

Install upgrade pip

or

Install Upgrade pip-break-system !

This will install the YOLOv8 (ultralytics) library and OpenCV library for working with images and video.

2. Verify the YOLO installation

To make sure YOLOv8 is installed correctly, run this command in the terminal:

yolo

If you see help with available YOLO commands, the installation was successful ✅

3. Downloading a pretrained model and first detection

YOLOv8 offers several versions of models – from lightweight to the most accurate ones. For the beginning, we recommend the yolov8n.pt (nano) model, which is very fast:

yolo detect predict model=yolov8n.pt source='0'

Yolo detect predict model Volov8n

This command will:

• download the yolov8n.pt model
• turn on your default camera
• display detected objects in real time 📸

Note: If you have multiple cameras, try source='1' or a higher number.

4. Creating your own Python script

For advanced usage, you can create your own Python script. Below is an example:

This script loads the model, starts the camera and in an infinite loop performs detection on each frame. The results are displayed in a separate window. Press q to exit the program.

The YOLOv8 model can recognize a wide range of objects – people, dogs, horses, cars, birds, skateboards and many others. The accuracy is significantly higher than older models such as MobileNet SSD.

Optional: switch to larger models

If you have more powerful hardware, you can use more accurate models:

• yolov8s.pt – small
• yolov8m.pt – medium
• yolov8l.pt – large
• yolov8x.pt – extra large (most accurate but slowest)

In the Python script, simply change:

model = YOLO("yolov8s.pt")

Frequently Asked Questions and Troubleshooting

⚠️ Warning “Could not initialize NNPACK!”

This is not an error. You just don’t have CPU optimizations for NNPACK → YOLO continues to work, only slightly slower.

Black screen after starting the camera

Possible reasons:

1️⃣ Camera is being used by another program
Close applications using the camera.

Check processes:

sudo fuser -v /dev/video0

2️⃣ Verify camera functionality using this script:

test-1

from ultralytics import YOLO
import cv2

model = YOLO("yolov8n.pt")
cap = cv2.VideoCapture(0)

if not cap.isOpened():
    print("❌ Camera failed to open")
    exit()

while True:
    ret, frame = cap.read()
    if not ret:
        print("❌ Camera returns no frame")
        break

    results = model.predict(source=frame, conf=0.5, verbose=False)
    annotated_frame = results[0].plot()

    cv2.imshow("YOLO Camera", annotated_frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

More tips for black screen:

• change 0 → 1 in VideoCapture()
• set resolution:

cap.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)

• add a small delay:

cv2.waitKey(10)

Older laptops ↴

If the hardware is weak, switch YOLO to CPU:

model.to('cpu')

Tips for speeding up:

• smaller frames (e.g. imgsz=320)
• CPU only (default)
• acceleration via TFLite / ONNX

Your final test file ✅

Designed for weak PCs with integrated GPU, e.g. 2nd generation i3. Should also work on Raspberry Pi 4.

# yolo_stream_light.py
from ultralytics import YOLO
import cv2
import time

model = YOLO("yolov8n.pt")
model.to("cpu")

cap = cv2.VideoCapture(0)
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 416)  # 416 often faster
# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 960)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 720)

if not cap.isOpened():
    print("Camera cannot be opened")
    exit()

print("Starting stream. Press q to exit.")
while True:
    ret, frame = cap.read()
    if not ret:
        print("Cannot load frame")
        break

    results_gen = model.predict(source=frame, conf=0.35, imgsz=416,
                               verbose=False, stream=True)

    try:
        res = next(results_gen)
    except StopIteration:
        res = None

    if res is not None:
        annotated = res.plot()
    else:
        annotated = frame

    cv2.imshow("YOLO light stream", annotated)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

Save the file with the .py extension and with no spaces in the name:
✅ yolo_stream_light.py

or Download

Open the terminal in the folder with your created program yolo_stream_light.py. Run the program in the terminal using the command: python3 yolo_stream_light.py

AI USB Camera: Real-Time Object Detection in Ubuntu Using YOLOv8