Introduction

This reporsitory is a simple example of YOLO real-time object detection.

It provides a basic Python program to capture video from camera and detect objects in real-time.

Requirements

To run this reporsitory, you need to follow the steps one by one.

As for the environment, I suggest using python>=3.8. (I use python=3.12)

How to use?

1. Clone this repository

git clone https://github.com/SiriusAhu/YOLO_Real_Time_Objection_Detection_Simple_Example

2. Create a virtual environment

If you're using Anaconda, you can create a virtual environment by:

conda create -n <env_name> python=3.12

Here, <env_name> is the name of your virtual environment. e.g. conda create -n yolo python=3.12

3. Activate your virtual environment

conda activate <env_name>

4. Install the required packages

You can choose to install the packages in two ways: CPU or GPU version.

4.1 I won't use GPU: Directly install ultralytics

UltraLytics provides a fantastic module called ultralytics, which contains almost all the required packages for using yolo.

So you can just install ultralytics by:

pip install ultralytics

4.2 I need GPU: Manually install torch, torchaudio, torchvision, then install ultralytics

Unfortunately, by testing, I found that ultralytics sometimes doesn't download GPU version of torch, torchaudio and torchvision correctly.

Instead you need to install them manually.

Follow the INSTALL PYTORCH part in the PyTorch official website to install torch, torchaudio and torchvision first.

The command is like this (but don't just copy and paste it, you need to check the website for the correct command):

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

4.3 Check the installation

You can use this command to check the installation:

pip freeze

You can check if torch, torchaudio, torchvision and ultralytics are GPU versions by checking whether they contain cuXXX in their names.

Then for further check if you install torch correctly, you can run the following code in Python:

import torch
print(torch.cuda.is_available())

If it returns True, then you have installed torch correctly.

5. Run the program

Just execute run.py with python (in your virtual environment)

python run.py

Note: Click Q to quit the program.

5.1 Arguments

Currently, only few arguments are provided:

1. -m or --model: Model to use. Enter 8n for YOLOv8n, 5s for YOLOv5s. (default: 8n)
2. -v or --verbose: Set to True for verbose output for model.predict(). (default: False)
3. -d or --device: Device to use. 0 for CPU, 1 for GPU. (default: 0)
4. -c or --camera: Camera to use. 0 for the first camera. Index is based on your system's camera index. Try to change this value if you meet errors. (default: 0)