plant_disease-mobilenetv2

🌿 Plant Disease Classification – MobileNetV2

High-Accuracy Deep Learning Model for Crop Disease Detection

Author: Daksh Goyal

🚀 Overview

This repository provides a lightweight, production-ready MobileNetV2 model trained on the PlantVillage augmented dataset for multi-class plant disease classification (38 classes).

The model is optimized for:

✔ Real-time inference ✔ Low-compute environments (mobile/edge devices) ✔ Web apps (Streamlit) ✔ Integration with farm-tech solutions

🧠 Model Summary

• Architecture: MobileNetV2
• Dataset: New Plant Diseases Dataset (Augmented) – PlantVillage
• Total Classes: 38
• Training Images: ~87,000 images (balanced across 38 classes)
• Validation Accuracy: 95%
• Loss: Cross Entropy
• Optimizer: Adam (LR = 0.001)
• Input Size: 224 × 224
• Format: PyTorch .pth

🌐 Live Demo

Try the fully deployed, interactive plant-disease detection app:

👉 Streamlit App: https://agriguard271005.streamlit.app/

Upload a leaf image → get: ✔ Disease name ✔ Confidence score ✔ Health status ✔ Severity level ✔ Recommended action

🤗 HuggingFace Model

The pretrained .pth model and configuration files are hosted here:

👉 HuggingFace Model Repo: https://huggingface.co/Daksh159/plant-disease-mobilenetv2

Available assets:

• mobilenetv2_plant.pth
• class_names.json
• README.md (this file)

🏗 Model Architecture (Training Setup)

Backbone: MobileNetV2

model = models.mobilenet_v2(pretrained=True)

for p in model.features.parameters():
    p.requires_grad = False  # freeze feature extractor

model.classifier[1] = nn.Sequential(
    nn.Dropout(0.2),
    nn.Linear(model.classifier[1].in_features, 38)
)

Training Loop

• 10 epochs
• Augmented inputs
• Adam optimizer
• Early stopping possible

📊 Supported Classes

Includes diseases & healthy leaves of:

• Apple
• Blueberry
• Cherry
• Corn
• Grape
• Peach
• Pepper
• Potato
• Raspberry
• Soybean
• Squash
• Strawberry
• Tomato
• Orange

(Full list of 38 classes included in class_names.json)

🖼 Inference Example

import torch
from torchvision import models, transforms
from PIL import Image

model = ...   # load your mobilenetv2_plant.pth
model.eval()

tf = transforms.Compose([
    transforms.Resize((224,224)),
    transforms.ToTensor(),
    transforms.Normalize([0.485,0.456,0.406],
                         [0.229,0.224,0.225])
])

img = Image.open("leaf.jpg").convert("RGB")
x = tf(img).unsqueeze(0)

with torch.no_grad():
    preds = model(x)
    probs = torch.softmax(preds, dim=1)[0]
    index = probs.argmax().item()

print("Predicted Class:", CLASS_NAMES[index])
print("Confidence:", float(probs[index]))

🔌 API / Integration Use Cases

This model can be integrated into:

🌱 Farming mobile apps 🌾 Agri-tech dashboards 💡 Real-time disease monitoring systems 📸 Drone-based crop scanning pipelines 🌍 IoT/edge deployment for field devices

🖥 Deployment

Available Deployments

🏁 Future Improvements

Planned enhancements:

• Grad-CAM heatmap explainability
• Faster ONNX / TensorRT export
• Support for multiple leaves in a single image
• Region-based crop disease localization
• Larger dataset fine-tuning

🙌 Acknowledgements

• PlantVillage Dataset
• PyTorch
• Kaggle
• Streamlit Cloud
• HuggingFace 🤗