---
license: apache-2.0
tags: [maize, ghana, agriculture, fall-armyworm, pytorch, computer-vision]
---
# Agricom Ghana Maize Disease Classifier

**Accuracy: 91.07%** on test set

Production model for detecting 7 maize diseases/pests in Ghana (Ashanti & Brong-Ahafo regions).

## Classes
1. Common Rust
2. Gray Leaf Spot  
3. Healthy
4. Maize Ear Rot
5. Fall Armyworm ⚠️ (triggers Emamectin Benzoate alerts)
6. Maize Stem Borer
7. Northern Leaf Blight

## Usage
```python
import torch
import torchvision.models as models
from torchvision import transforms
from PIL import Image

# Load model
checkpoint = torch.load('model.pth')
model = models.resnet18()
model.fc = torch.nn.Linear(512, 7)
model.load_state_dict(checkpoint['state_dict'])
model.eval()

# Preprocess image
transform = transforms.Compose([
    transforms.Resize(256),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])

# Predict
img = Image.open('maize_leaf.jpg').convert('RGB')
img_tensor = transform(img).unsqueeze(0)

with torch.no_grad():
    output = model(img_tensor)
    probabilities = torch.nn.functional.softmax(output[0], dim=0)
    pred_class = output.argmax(1).item()

print(f"Prediction: {checkpoint['class_names'][pred_class]}")
print(f"Confidence: {probabilities[pred_class]:.2%}")
```

## Model Details
- **Architecture**: ResNet-18 (torchvision)
- **Input**: 224×224 RGB images
- **Training**: Transfer learning from ImageNet
- **Framework**: PyTorch