Running YOLO V7 on Windows 10/11

Requirements

• Hardware: Laptop/PC, GPU (optional but recommended), Internet connection
• Software: Python 3, Conda

Installing Conda

1. Download Conda: Go to the Conda website and choose Anaconda from the versions (Anaconda, Miniconda, Anaconda Enterprise)
2. Install:
   • Click the Windows download button (600MB file)
   • Agree to terms, select installation type (Just Me), and add Anaconda to the PATH environment variable
   • Click Install

Creating Conda Environment

1. Create Environment:

   • Open Command Prompt
   • Type: conda create --name any_env_name
   • Activate with: conda activate any_env_name

2. Clone YOLO V7 Repository:

   • Use git: git clone [YOLO V7 link]

3. Install Required Packages:

   • Use pip: pip install -r requirements.txt

Downloading Pre-trained Weights

1. Access Weights:
   • Available on YOLO V7 GitHub repository
   • Look for YOLO v7.pt under the testing section
2. Save Weights:
   • Place them in the YOLO V7 folder

Detection on Images using GPU

1. Install CUDA:
   • Install NVIDIA CUDA toolkit compatible with your GPU and system
   • Download and install from NVIDIA website
2. GPU Setup:
   • Ensure python detect.py --weights [path] --conf 0.5 --img-size [size] --source [image file] --device 0
   • Run the detection code
   • Check result in YOLO V7 runs/detect/experiment folder

Detection on Images using CPU

1. Change Device Setting:
   • Set device to CPU --device cpu
   • Detection will take longer
   • Results saved similarly in runs/detect/experiment folder

Detection on Videos

1. Source Setup:
   • Replace source from image to video file path --source [video file]
   • Process similarly as with images
2. Run the Model:
   • Check results in runs/detect/experiment folder

Detection via Webcam

1. Setup Source:
   • Change source to 0 for inbuilt webcam
   • Change to 1 for an external webcam

YOLO V7 Custom Training (Brief Mention)

• Tutorial based on official YOLO V7 repository instructions.