YOLO Keyboard Key Detection

A project for detecting keyboard keys using YOLOv8, optimized for NVIDIA RTX 4060 (8GB VRAM).

---

🎥 Demo Video

Watch the full demonstration: YOLO Keyboard Key Detection in Action

---

🚀 Features

• Train YOLOv8 models for keyboard key detection
• Inference on images and real-time webcam streams
• Easy integration with Roboflow datasets
• Optimized for consumer GPUs

---

📁 Project Structure

.
├── scripts/
│   ├── train_keyboard_detector.py
│   └── test_inference.py
├── data/
│   └── keyboard_dataset/         # (excluded from repo, see below)
├── results/                      # (excluded from repo)
├── requirements.txt
├── .gitignore
└── README.md

---

🗂️ Dataset

Dataset is NOT included in this repository.

• Download your dataset from Roboflow (export in YOLOv8 format).
• Place it in:
  data/keyboard_dataset/
• Your folder should look like:

  data/keyboard_dataset/
    images/
      train/
      valid/
      test/
    labels/
      train/
      valid/
      test/
    data.yaml
  

---

⚙️ Setup

1. Clone this repo:

   git clone https://github.com/anishk85/yolo-keyboard-key-detection.git
   cd yolo_keyboard_project

2. Create and activate a Python environment (optional but recommended):

   python3 -m venv yolo_keyboard_env
   source yolo_keyboard_env/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

---

🏋️‍♂️ Training

Train your YOLO model with:

python3 scripts/train_keyboard_detector.py

• Adjust parameters in train_keyboard_detector.py as needed.

---

Before training or inference, you can verify that your environment and dependencies are set up correctly:

python3 scripts/verify_installation.py

This script will:

• Check your system info (CPU, RAM, Python version)
• Check GPU and CUDA availability
• Check all required Python libraries and their versions
• Test YOLO model loading and inference (CPU and GPU)
• Test GPU memory usage and recommend a batch size

If all tests pass, your system is ready for YOLO keyboard detection!

---

🔍 Inference

On a single image:

python3 scripts/test_inference.py --model <path_to_best.pt> --image <path_to_image>

With your webcam:

python3 scripts/test_inference.py --model <path_to_best.pt> --webcam

List available cameras:

python3 scripts/test_inference.py --list-cameras

---

📝 Notes

• Do NOT commit large datasets or model weights.
  See .gitignore for excluded files.
• For best results, use a GPU.
• For dataset details and annotation format, see Roboflow documentation

---

📄 License

This project is licensed under the MIT License.

---

🙏 Acknowledgements

• Ultralytics YOLO
• Roboflow