AgriVision: AI Crop Diagnostics

• 

  Gradio

• 

  Automated Accuracy Comparison

• 

  Accuracy for transfer learning

• 

  Confusion Matrix

• 

  Automated Accuracy Comparison

🍇 Inspiration

Vineyard diseases like Black Rot and Esca cause significant economic losses annually. As an AI student, I wanted to build a high-precision tool that empowers farmers with instant, expert-level diagnostics using only a smartphone camera.

🛠️ What it does

VitiVision identifies three categories of grape leaf health: Black Rot, Esca (Black Measles), and Healthy. Users simply upload a photo to our Gradio-powered web interface and receive a diagnostic result in seconds with a confidence score.

🚀 How I built it

• Model: Utilized MobileNetV2 Transfer Learning to leverage pre-trained visual hierarchies.
• Dataset: Trained on a balanced dataset of 1,500 grape leaf images.
• Optimization: Achieved a peak validation accuracy of 99.33% and a stable average of 96.33%.
• Deployment: Built a user-friendly interface using Gradio for real-time inference.

📈 Challenges & Accomplishments

The biggest challenge was preventing overfitting with a limited dataset. I successfully implemented Dropout layers and Data Augmentation to ensure the model generalizes well to real-world vineyard photos.

Built With

• gradio
• keras
• matplotlib
• mobilenetv2
• python
• scikit-learn
• tensorflow