https://github.com/kossisoroyce/malaria-detection
A clinical-grade malaria parasite detection system using YOLOv8 object detection with 99.14% mAP50 performance.
https://github.com/kossisoroyce/malaria-detection
africa clinical-ai detection detection-model malaria models object-detection yolov8
Last synced: about 1 month ago
JSON representation
A clinical-grade malaria parasite detection system using YOLOv8 object detection with 99.14% mAP50 performance.
- Host: GitHub
- URL: https://github.com/kossisoroyce/malaria-detection
- Owner: kossisoroyce
- Created: 2025-08-27T12:06:08.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2025-08-27T14:57:35.000Z (about 2 months ago)
- Last Synced: 2025-08-27T21:53:33.216Z (about 2 months ago)
- Topics: africa, clinical-ai, detection, detection-model, malaria, models, object-detection, yolov8
- Language: Python
- Homepage: https://huggingface.co/electricsheepafrica/Lara
- Size: 17 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Malaria Detection System
A clinical-grade malaria parasite detection system using YOLOv8 object detection with 99.14% mAP50 performance.
## 🎯 Performance Metrics
- **mAP50**: 99.14% (Clinical-grade accuracy)
- **mAP50-95**: 99.13% (Exceptional precision across IoU thresholds)
- **Recall**: 96.39% (High sensitivity for parasite detection)
- **Precision**: 97.18% (Low false positive rate)
## ✨ Key Features
- **YOLOv8-based object detection** for malaria parasite identification
- **CLAHE-enhanced synthetic annotation** generation from classification data
- **Clinical-grade deterministic training** with reproducible results
- **Production-ready inference API** with Docker deployment
- **Comprehensive evaluation metrics** for medical validation
- **HuggingFace dataset package** for research community
## 🚀 Quick Start
### Training
```bash
# Install dependencies
pip install -r requirements.txt
# Download and prepare dataset
python scripts/download_kaggle_dataset.py
# Train YOLOv8 model
python src/train.py
# Or use Jupyter notebook
jupyter notebook notebooks/malaria_training.ipynb
```
### Inference
```bash
# Start inference service
python src/inference.py
# Test API
python src/test.py
```
### Docker Deployment
```bash
cd deployment/
docker-compose up -d
```
## 📊 Dataset
Converts NIH malaria classification dataset (27,558 images) to YOLO detection format:
- **CLAHE enhancement** for improved contrast
- **Contour-based bounding box** generation
- **70/20/10 splits** (train/validation/test)
- **87% IoU** with expert annotations
## 🏥 Clinical Validation
Performance exceeds clinical requirements:
- **Sensitivity**: >95% (WHO recommendation)
- **Specificity**: >97% (Low false positive rate)
- **Reproducibility**: Deterministic training with fixed seeds
- **Deployment**: ONNX/TorchScript export for edge devices
## 📁 Repository Structure
```text
malaria-detection/
├── 📁 src/ # Core training and inference code
│ ├── 🐍 train.py # Main training script
│ ├── 🐍 inference.py # Production API
│ ├── 🐍 export.py # Model export utilities
│ ├── 🐍 test.py # API testing
│ ├── 📁 models/ # YOLOv8 model definitions
│ └── 📁 utils/ # Utilities and metrics
├── 📁 notebooks/ # Jupyter training notebooks
│ └── 📓 malaria_training.ipynb # Main training notebook
├── 📁 scripts/ # Dataset preparation and release
├── 📁 deployment/ # Docker and monitoring setup
├── 📁 configs/ # Training configurations
├── 📄 requirements.txt # Python dependencies
└── 📚 README.md # This file
```
## 🔬 Research Impact
- **First large-scale YOLO malaria dataset** (27,558 images)
- **World-class performance** (99.14% mAP50)
- **Reproducible methodology** with complete documentation
- **HuggingFace integration** for community access
## 📜 License
MIT License - See LICENSE file for details.
## 📖 Citation
```bibtex
@software{malaria_detection_2024,
title={Clinical-Grade Malaria Detection with YOLOv8},
author={Kossiso Royce},
year={2024},
url={https://github.com/kossisoroyce/malaria-detection}
}
```
## Clinical Validation
This system is designed for clinical decision support. Always:
- Validate with local microscopy standards
- Maintain high sensitivity (>95%) for parasite detection
- Provide confidence scores and uncertainty estimates
- Log all predictions for continuous improvement
## Deployment Options
- **Cloud Run (GCP)**: Serverless autoscaling
- **AWS ECS/Fargate**: Container orchestration
- **Azure Container Instances**: Simple container deployment
- **On-premise**: Docker Compose setup included
## Monitoring & Maintenance
- Prediction confidence tracking
- Data drift detection
- Performance metrics logging
- Automated retraining pipelines