🔬 YOLOv5 — Microparticle Detection & Classification
Automated detection and classification of microplastics from microscopy images. Developed for industrial deployment at IZONICS — achieving >90% accuracy in production.
📌 Model Summary
| Property | Details |
|---|---|
| 🏗️ Architecture | YOLOv5 |
| 🎯 Task | Object Detection + Classification |
| 📊 Accuracy | >90% in production |
| ⚡ Speedup | 30% faster than classical image processing pipeline |
| 🏭 Deployment | Fully industrialized — prototype to production |
| ⚙️ Framework | PyTorch |
| 📜 License | MIT |
🧠 What This Model Does
This model detects and classifies microplastic particles in microscopy images, enabling automated environmental monitoring and industrial quality control.
It was developed at IZONICS as a complete refactoring of an existing classical computer vision pipeline into a deep learning architecture — resulting in higher reliability, better accuracy, and significantly faster inference in real production environments.
Origin: Extended from a Master's thesis (VIBOT program, Université de Bourgogne) into a fully deployed industrial product.
📈 Performance
| Metric | Score |
|---|---|
| Production accuracy | >90% |
| Processing speedup vs. classical pipeline | 30% |
| Deployment status | ✅ Fully industrialized |
🚀 How to Use
import torch
# Load the model
model = torch.hub.load("ultralytics/yolov5", "custom", path="best.pt")
model.conf = 0.35 # confidence threshold
# Run inference
results = model("microscopy_image.jpg")
# Show results
results.show()
# Get detections as dataframe
df = results.pandas().xyxy[0]
print(df)
🗂️ Training Data
- Domain: Environmental microscopy images of microplastic particles
- Task: Multi-class detection and classification of particle types
- Augmentation: Custom augmentation tailored for microscopy image characteristics
⚠️ Limitations
- Optimized for specific microscopy imaging conditions used at IZONICS
- Performance may vary on microscopy setups with different magnification or lighting
- Not validated for clinical or regulatory use
🔗 Related Resources
👤 Author
Martin Badrous — Computer Vision & Deep Learning Engineer
- Downloads last month
- 2