Kathir56/plant-disease-tamilnadu

🌿 Plant Disease Classifier for Tamil Nadu Agriculture

MobileNetV2 fine-tuned on the PlantVillage dataset for classifying 38 plant diseases relevant to Indian agriculture.

Model Details

• Architecture: MobileNetV2 (2.3M params) — optimized for mobile/edge deployment
• Base model: google/mobilenet_v2_1.0_224
• Training data: PlantVillage (8,000 stratified train / 10,849 full test)
• Classes: 38 disease categories

Performance

Metric	Value
Accuracy	97.31%
F1 (weighted)	97.31%

Training Configuration

• Epochs: 5
• Batch size: 32
• Learning rate: 0.0003 (cosine schedule with warmup)
• Label smoothing: 0.1
• Weight decay: 0.01
• Augmentation: RandomResizedCrop, HFlip, VFlip, Rotation(15°), ColorJitter

Tamil Nadu Relevant Classes ⭐

This model covers crops commonly grown in Tamil Nadu:

• Tomato (10 classes): Bacterial spot, Early/Late blight, Leaf Mold, Septoria, Spider mites, Target Spot, Yellow Leaf Curl, Mosaic Virus, Healthy
• Corn/Maize (4): Cercospora, Common Rust, Northern Leaf Blight, Healthy
• Grape (4): Black Rot, Esca, Leaf Blight, Healthy
• Pepper/Chilli (2): Bacterial Spot, Healthy
• Potato (3): Early Blight, Late Blight, Healthy
• Orange/Citrus (1): Huanglongbing (Citrus Greening)

Usage

from transformers import pipeline

classifier = pipeline("image-classification", model="Kathir56/plant-disease-tamilnadu")
result = classifier("path/to/leaf_image.jpg")
print(result)
# Example output: [{'label': 'Tomato___Late_blight', 'score': 0.95}]

Or with PIL Image:

from transformers import MobileNetV2ForImageClassification, MobileNetV2ImageProcessor
from PIL import Image

processor = MobileNetV2ImageProcessor.from_pretrained("Kathir56/plant-disease-tamilnadu")
model = MobileNetV2ForImageClassification.from_pretrained("Kathir56/plant-disease-tamilnadu")

image = Image.open("leaf.jpg")
inputs = processor(images=image, return_tensors="pt")
outputs = model(**inputs)
predicted_class = outputs.logits.argmax(-1).item()
print(model.config.id2label[predicted_class])

All 38 Classes

• Apple___Apple_scab
• Apple___Black_rot
• Apple___Cedar_apple_rust
• Apple___healthy
• Blueberry___healthy
• Cherry_(including_sour)___Powdery_mildew
• Cherry_(including_sour)___healthy
• Corn_(maize)___Cercospora_leaf_spot Gray_leaf_spot
• Corn_(maize)__Common_rust
• Corn_(maize)___Northern_Leaf_Blight
• Corn_(maize)___healthy
• Grape___Black_rot
• Grape___Esca_(Black_Measles)
• Grape___Leaf_blight_(Isariopsis_Leaf_Spot)
• Grape___healthy
• Orange___Haunglongbing_(Citrus_greening)
• Peach___Bacterial_spot
• Peach___healthy
• Pepper,_bell___Bacterial_spot
• Pepper,_bell___healthy
• Potato___Early_blight
• Potato___Late_blight
• Potato___healthy
• Raspberry___healthy
• Soybean___healthy
• Squash___Powdery_mildew
• Strawberry___Leaf_scorch
• Strawberry___healthy
• Tomato___Bacterial_spot
• Tomato___Early_blight
• Tomato___Late_blight
• Tomato___Leaf_Mold
• Tomato___Septoria_leaf_spot
• Tomato___Spider_mites Two-spotted_spider_mite
• Tomato___Target_Spot
• Tomato___Tomato_Yellow_Leaf_Curl_Virus
• Tomato___Tomato_mosaic_virus
• Tomato___healthy

Recommended Next Steps

For better Tamil Nadu crop coverage, consider:

1. Paddy Doctor dataset — rice diseases specific to Tamil Nadu
2. Sugarcane disease datasets — major TN crop
3. Banana, Coconut, Tea disease data
4. Fine-tune with Tamil Nadu field imagery for domain adaptation
5. Use as a teacher model for knowledge distillation to even smaller architectures

Citation

@article{hughes2015open,
  title={An open access repository of images on plant health to enable development of mobile disease diagnostics},
  author={Hughes, David P and Salathe, Marcel},
  journal={arXiv preprint arXiv:1511.08060},
  year={2015}
}