Mobile Phone Detection using YOLOv5

This project implements a real-time mobile phone detection system using the YOLOv5 object detection architecture. It provides scripts for training a custom model, running real-time detection on a webcam, and validating model performance.

🚀 Features

• Real-time Detection: Detect mobile phones in real-time using your webcam.
• Custom Training: Tools to fine-tune YOLOv5 on your own mobile phone dataset.
• Validation: Evaluate model performance with precision/recall metrics.
• Model Export: Export trained models to ONNX, TorchScript, or TFLite for deployment.

📂 Project Structure

• train_mobile_phone.py: The main utility script for training, validating, and detecting.
• stop_on_phone_detect.py: A standalone script that runs detection using the pre-trained COCO model (class 67: cell phone) and provides a summary.
• run_detection.py: A shortcut to launch the real-time detection mode of the main script.
• mobile_phone_dataset/: Directory structure for your custom dataset (images and labels).
• runs/: Stores training results (weights, confusion matrices, loss plots).

🛠️ Installation

1. Clone the repository:

   git clone https://github.com/wareeshayy/CV_MobilePhoneDetection.git
   cd CV_MobilePhoneDetection

2. Set up the environment: Create a virtual environment (recommended) and install dependencies.

   # Windows
   python -m venv venv
   .\venv\Scripts\activate
   
   # Linux/Mac
   python3 -m venv venv
   source venv/bin/activate

3. Install YOLOv5 Dependencies: The scripts will automatically clone the YOLOv5 repository if missing, but you need to install its requirements.

   pip install -r yolov5/requirements.txt

⚡ Quick Start

To see mobile phone detection in action immediately using the standard pre-trained YOLOv5s model (trained on COCO dataset):

python stop_on_phone_detect.py

This will:

1. Open your webcam.
2. Detect mobile phones (class 'cell phone').
3. Stop automatically after 300 frames.
4. Print a detection summary (total phones detected, frames with detections).

🏋️ Training on Custom Data

To train a model specifically for your environment or better accuracy:

1. Prepare your Dataset: Organize your images and labels in the mobile_phone_dataset directory:

   mobile_phone_dataset/
   ├── images/
   │   ├── train/  (training images)
   │   └── val/    (validation images)
   └── labels/
       ├── train/  (YOLO format txt files)
       └── val/    (YOLO format txt files)
   

2. Run the Training Script:

   python train_mobile_phone.py

3. Select Option 1 to start training.

   • You can configure epochs, batch size, and model size in the script or follow the prompts.

🔍 Running Detection with Custom Model

Once you have trained your own model, you can run inference using:

python run_detection.py

Alternatively, run train_mobile_phone.py and select Option 3.

📊 Evaluation & Export

• Validation: Run train_mobile_phone.py and select Option 2 to see Mean Average Precision (mAP) scores.
• Export: Run train_mobile_phone.py and select Option 4 to convert your model for mobile or edge deployment (ONNX, TFLite, etc.).

📝 License

This project is open source and available under the MIT License.