Machine Vision Technology Development

Experiment 4 - YOLOv5 Object Detection

Experiment steps:

  1. Need to install python3, opencv, conda environment. (Can skip if already installed, can refer to experiments 1-3 for environment installation process)
  2. Clone YOLOv5 model, enter command in terminal: git clone https://github.com/ultralytics/yolov5
TOOL
  1. cp -r /home/sunrise/yolov5 /home/sunrise/OPENCV/ # Copy yolov5 package to function package
cd OPENCV
pip install -r /home/sunrise/OPENCV/requirements_yolov5_torch.txt

(Install minimum dependencies for YOLOv5 to run)

TOOL

  1. source rdkx5_vision_env/bin/activate # Activate virtual environment

  2. Install related packages:

In OPENCV directory

First upgrade pip: python -m pip install --upgrade pip

Run the following command:

pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cpu)
pip install ultralytics
pip install pandas psutil thop scipy
python -m pip install tqdm

It is recommended to install together as they may be used later: python -m pip install pandas psutil thop pillow pyyaml requests matplotlib seaborn

(The following steps can be skipped first, try running the sample file first, if version dependency is too low and cannot run the model, then update:

Upgrade basic installation tools: python -m pip install -U pip wheel setuptools==70.0.0

Install/upgrade missing dependencies: python -m pip install -U gitpython pillow==10.3.0

)

  1. cd yolov5 # Enter package
  2. python detect.py --weights yolov5s.pt --source 0 # Run camera version sample file, ensure camera is connected properly

Terminal prints as follows:

TOOLTOOL

If no camera is available, you can choose local image or video import:

  • Use local image for quick verification

  • python detect.py --weights yolov5s.pt --source path\to\image.jpg # Change to image path after source

  • Use video file for verification

  • python detect.py --weights yolov5s.pt --source path\to\video.mp4 # Change to video path after source