https://github.com/mosesab/yolov11-drowsiness-detection
This repo contains the fine-tune implementation of a drowsy detection system, using a YOLO v11 base model.
https://github.com/mosesab/yolov11-drowsiness-detection
computer-vision drowsiness-detection drowsiness-detection-python fine-tuning yolo yolov11
Last synced: 4 months ago
JSON representation
This repo contains the fine-tune implementation of a drowsy detection system, using a YOLO v11 base model.
- Host: GitHub
- URL: https://github.com/mosesab/yolov11-drowsiness-detection
- Owner: mosesab
- License: mit
- Created: 2024-12-30T22:05:42.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2024-12-30T22:52:45.000Z (6 months ago)
- Last Synced: 2024-12-30T23:18:02.983Z (6 months ago)
- Topics: computer-vision, drowsiness-detection, drowsiness-detection-python, fine-tuning, yolo, yolov11
- Language: Jupyter Notebook
- Homepage:
- Size: 6.04 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Drowsy Detection with YOLOv11
This repository contains the full implementation of a drowsy detection system, leveraging the YOLO (You Only Look Once) object detection algorithm fine-tuned for the "awake" and "drowsy" datasets. The project uses the Python `ultralytics` library for training and inference.
## Table of Contents
- [Dataset](#dataset)
- [Files and Structure](#files-and-structure)
- [Training Process Overview](#training-process-overview)
- [Dependencies and Requirements](#dependencies-and-requirements)
- [Key Results](#key-results)
- [Citation](#citation)
---
## Dataset
The dataset used for training can be downloaded from the following link:
[Complete YOLO Drowsy Images Dataset](https://firebasestorage.googleapis.com/v0/b/electora-8c1d6.appspot.com/o/Complete%20YOLO%20Drowsy%20Images%20Dataset.zip?alt=media)
The dataset includes annotated images in YOLO format, with labels specifying the "awake" and "drowsy" states.
---
## Files and Structure
### Repository Structure:
```plaintext
│ notebook.ipynb # Initial training session
│ notebook2.ipynb # Resumed training session
├─training_and_testing_files: # Outputs and intermediate results
│ args.yaml
│ confusion_matrix.png
│ confusion_matrix_normalized.png
│ F1_curve.png
│ labels.jpg
│ labels_correlogram.jpg
│ PR_curve.png
│ P_curve.png
│ results.csv
│ results.png
│ results1.png
│ results2.png
│ results3.png
│ results4.png
│ results5.png
│ R_curve.png
│ train_batch*.jpg
│ val_batch*_labels.jpg
│ val_batch*_pred.jpg
├─.ipynb_checkpoints: # Checkpointed files
└─pics: # Selected visualization outputs
confusion_matrix.png
F1_curve.png
labels.jpg
train_batch34880.jpg
val_batch1_pred.jpg
```
---
## Training Process Overview
1. **Dataset Loading:**
- YOLO utilizes a `yaml` file to define the paths for training and validation datasets and their corresponding class labels.
2. **Key Parameters for Training:**
- **Epochs:** The model was trained for 50 epochs to balance learning and avoiding overfitting.
- **Image Size:** All input images were resized to 640x640 pixels to match YOLO's expected square input format.
- **Batch Size:** A batch size of 16 was used to optimize training efficiency.
3. **Hardware and Framework:**
- **Framework:** PyTorch backend via the `ultralytics` library.
- **Hardware:** Training and inference were performed on an Nvidia RTX 3080 GPU.
---
## Dependencies and Requirements
- Python 3.8+
- Key Python libraries:
- `ultralytics` for YOLO-based training and inference
- `torch` and `torchvision` for PyTorch operations
- `numpy`, `matplotlib`, and `pandas` for data processing and visualization
- GPU with CUDA support (recommended: Nvidia RTX 3080 or higher)
Install dependencies using the following command:
```bash
pip install ultralytics torch torchvision numpy matplotlib pandas
```
---
## Key Results
- **Performance Metrics:**
- Confusion Matrix: `confusion_matrix.png` and `confusion_matrix_normalized.png`
- Precision-Recall (PR) Curve: `PR_curve.png`
- F1 Curve: `F1_curve.png`
- Results stored in `results.csv` and visualized in `results*.png`
- **Visualizations:**
- Training Batches: `train_batch*.jpg`
- Validation Predictions: `val_batch*_pred.jpg`
---
## Citation
If you use this repository in your research, please cite it using the following DOI link:
[DOI Placeholder]()
Feel free to open issues or pull requests if you encounter any problems or have suggestions for improvement.
---
**License:** This project is licensed under the MIT License. See `LICENSE` for details.