|
|
--- |
|
|
license: agpl-3.0 |
|
|
library_name: ultralytics |
|
|
pipeline_tag: object-detection |
|
|
tags: |
|
|
- yolo |
|
|
- ultralytics |
|
|
- yolov11 |
|
|
- object-detection |
|
|
- ppe-detection |
|
|
- computer-vision |
|
|
- safety |
|
|
datasets: |
|
|
- custom |
|
|
--- |
|
|
|
|
|
 |
|
|
|
|
|
# PPE Detection with YOLOv11 |
|
|
This model is a specialized version of YOLOv11, fine-tuned to detect Personal Protective Equipment (PPE) in industrial and construction environments. It is designed to enhance occupational safety by monitoring the use of helmets and safety vests in real-time. |
|
|
|
|
|
|
|
|
## 🚀 Quick Start (Usage) |
|
|
You don't need to download the weights manually. You can load and run the model directly using the Python code below: |
|
|
|
|
|
```python |
|
|
from ultralytics import YOLO |
|
|
from huggingface_hub import hf_hub_download |
|
|
|
|
|
# Downloading and loading the model |
|
|
model_path = hf_hub_download(repo_id="melihuzunoglu/ppe-detection", filename="best.pt") |
|
|
model = YOLO(model_path) |
|
|
|
|
|
# Run inference |
|
|
results = model.predict(source="construction_site.jpg", conf=0.25, save=True) |
|
|
``` |
|
|
|
|
|
|
|
|
## ✅ Supported Classes (Labels) |
|
|
The model follows a hierarchical detection strategy for higher accuracy: |
|
|
```python |
|
|
Human: Detects the person/worker as the primary anchor. |
|
|
|
|
|
Helmet: Detects safety helmets worn on the head. |
|
|
|
|
|
No-Helmet: Specifically identifies heads without safety helmets. |
|
|
|
|
|
Vest: Detects high-visibility safety vests. |
|
|
``` |
|
|
Note: For "No-Vest" detection, the logic is based on the absence of a vest label within a detected human bounding box. |
|
|
|
|
|
|
|
|
|
|
|
## 📊 Model Information |
|
|
```python |
|
|
Architecture: YOLOv11 (Ultralytics) |
|
|
|
|
|
Task: Object Detection (PPE Compliance) |
|
|
|
|
|
Input Resolution: 640x640 pixels |
|
|
|
|
|
Training Strategy: Hierarchical labeling (Human as anchor) to reduce false negatives in complex backgrounds. |
|
|
``` |
|
|
|
|
|
|
|
|
## 🎯 Target Applications |
|
|
```python |
|
|
Construction Sites: Real-time monitoring of helmet and vest compliance. |
|
|
|
|
|
Industrial Plants: Automated safety audits for manufacturing floors. |
|
|
|
|
|
Mining & Logistics: Ensuring worker visibility and protection in hazardous zones. |
|
|
|
|
|
Safety Training: Providing visual feedback during safety drills. |
|
|
``` |
|
|
|
|
|
|
|
|
## 🛠 Training Details |
|
|
The model was trained using the Ultralytics framework. The dataset was meticulously curated and pre-processed via Roboflow, utilizing a hierarchical approach where equipment is detected in relation to the human figure. This ensures that the model focuses on the equipment's visual features while maintaining spatial awareness of the worker. |
|
|
|
|
|
|
|
|
## 👤 Developer |
|
|
Author: Melih Uzunoğlu [Linkedin](https://www.linkedin.com/in/melih-uzunoglu/) |
|
|
|
|
|
Framework: Ultralytics YOLOv11 |
|
|
|
|
|
Dataset Source: Roboflow |
|
|
|
|
|
|
|
|
### Disclaimer |
|
|
This model is developed for educational and research purposes. While it provides high-accuracy detections, it should be used as a supplementary tool alongside professional safety inspections in critical workplace environments. |
|
|
|
