|
|
--- |
|
|
tags: |
|
|
- image-classification |
|
|
- pytorch |
|
|
- computer-vision |
|
|
- lanyard-detection |
|
|
- sdg-16 |
|
|
license: mit |
|
|
datasets: |
|
|
- custom |
|
|
metrics: |
|
|
- accuracy |
|
|
library_name: pytorch |
|
|
--- |
|
|
|
|
|
# Lanyard Detector - School Safety AI |
|
|
|
|
|
This model detects whether a person is wearing a lanyard/ID badge around their neck. |
|
|
|
|
|
## 🎯 Use Case |
|
|
|
|
|
**Problem:** Manual lanyard checking at school entrances is time-consuming. |
|
|
|
|
|
**Solution:** Automated computer vision system for real-time lanyard detection. |
|
|
|
|
|
**SDG Impact:** SDG 16 (Peace, Justice & Strong Institutions) - School safety |
|
|
|
|
|
## 📊 Model Details |
|
|
|
|
|
- **Architecture:** MobileNetV2 (Transfer Learning) |
|
|
- **Input:** RGB images (224x224) |
|
|
- **Output:** Binary classification |
|
|
- Class 0: `has_lanyard` |
|
|
- Class 1: `no_lanyard` |
|
|
- **Framework:** PyTorch |
|
|
|
|
|
## 🚀 Quick Start |
|
|
|
|
|
```python |
|
|
import torch |
|
|
from torchvision import transforms, models |
|
|
from PIL import Image |
|
|
|
|
|
# Load model |
|
|
checkpoint = torch.load('pytorch_model.pth', map_location='cpu') |
|
|
model = models.mobilenet_v2() |
|
|
model.classifier[1] = torch.nn.Linear(1280, 2) |
|
|
model.load_state_dict(checkpoint['model_state_dict']) |
|
|
model.eval() |
|
|
|
|
|
# Preprocess |
|
|
transform = transforms.Compose([ |
|
|
transforms.Resize((224, 224)), |
|
|
transforms.ToTensor(), |
|
|
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) |
|
|
]) |
|
|
|
|
|
# Predict |
|
|
img = Image.open('test.jpg') |
|
|
img_tensor = transform(img).unsqueeze(0) |
|
|
|
|
|
with torch.no_grad(): |
|
|
output = model(img_tensor) |
|
|
probs = torch.softmax(output, dim=1) |
|
|
pred_class = output.argmax(1).item() |
|
|
|
|
|
classes = ['has_lanyard', 'no_lanyard'] |
|
|
print(f"Prediction: {classes[pred_class]}") |
|
|
print(f"Confidence: {probs[0][pred_class]*100:.1f}%") |
|
|
``` |
|
|
|
|
|
## 📝 Training Details |
|
|
|
|
|
- Custom dataset with data augmentation |
|
|
- 70% train / 15% val / 15% test split |
|
|
- Adam optimizer (lr=0.001) |
|
|
- 15 epochs |
|
|
|
|
|
## ⚡ Deployment |
|
|
|
|
|
Deploy via: |
|
|
- Featherless AI (Hugging Face partner) |
|
|
- ONNX Runtime |
|
|
- TorchScript |
|
|
- Edge devices |
|
|
|
|
|
## ⚠️ Limitations |
|
|
|
|
|
- Best with frontal/slight side angles |
|
|
- Requires adequate lighting |
|
|
- May struggle with very small/distant lanyards |
|
|
|
|
|
## 📄 License |
|
|
|
|
|
MIT License |
|
|
|
|
|
--- |
|
|
|
|
|
**Developed for:** Goals in Code Hackathon 2026 |
|
|
**SDG:** 16 - Peace, Justice & Strong Institutions |
|
|
|
