Staff Detection

This project applies a YOLOv8-based deep learning model to detect staff members video footage using bounding box logic. It was trained on custom-labeled data with staff and tag annotations, and deployed with post-processing utilities for accurate staff detection in surveillance videos.

Model Training in Colab:

https://colab.research.google.com/drive/1dIbm6Ld0lTjNXL67gQj5yiMSF4YMX_ye?usp=sharing

🔧 Features

• ✅ Trained YOLOv8 model for object detection
• ✅ Bounding box comparison: staff vs staff_tag
• ✅ Coordinate extraction from detection output
• ✅ Easy deployment on local machine (GPU supported)
• ✅ Google Colab notebook for reproducible training

📁 Project Structure

├── README.md          <- The top-level README for developers using this project.
├── docs               <- Documentations
│   └── Staff_Detect_Report.docx     <- Report of Model.
│   
├── my_model           <- Directory of model, outputs
│   ├── train              <- Model training reports.
│   ├── my_model.pt        <- Best weights 
│   └── yolo_detect.py     <- Detection utils
│ 
└── .gitignore

Deploying locally

Refference

    git clone https://github.com/EASONTAN03/FootfallCam_staff_detect

1. Unzip <my_model.zip>

2. Create python env:

    conda create --name yolo-env1 python=3.12 -y
   
    conda activate yolo-env1
   

3. Install requirements:

    pip install ultralytics
   

   Install Nvidia GPU-enabled version of PyTorch by issuing the following command:

    pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124
   

4. Locate the file: cd my_model

5. Modify detection utils file created by Evan Juras, EJ Technology Consultants. [yolo_detect.py]

• Add coordinates extraction function.
• Add bouding box comparison for staff detection: if box of staff_tag is inside box of staff then staff detected.

8. Detect objects for video:

    python yolo_detect.py --model my_model.pt --source ../sample.mp4 --resolution 640x640 --coordinates True
   

Demo of running the detection

Demo_staff_detect