Tensor Crop Tracker (TCT) AI

Introducing TCT AI: Your pocket plant tracking system. Our app revolutionizes crop management by diagnosing plant diseases and providing tailored, eco-friendly treatment recommendations.

Comment

Inspiration

Farmers worldwide have long used pesticides and harmful chemicals to treat their crops, leading to damaging effects on human health and the environment. Many of these farmers do not know about the existence of natural alternatives, or haven't stopped to consider them.

What it does

Tensor Crop Tracker (TCT) AI is a disease detection system that uses artificial intelligence and computer vision to analyze plant health. By using a mobile device to capture photos of crops, TCT AI can:

  • Diagnose plant diseases within minutes
  • Provide eco-friendly recommendations to cure these diseases
  • Recommend measures for disease prevention in the future

How we built it

We developed TCT AI by integrating multiple technologies together:

  • A convolutional neural network (CNN) trained on a large, diverse dataset of plant images for disease classification
  • Accessing UIUC Chat with API keys (Qwen2.5-VL-72B-Instruct model) for recommending treatments
  • A mobile application built with SwiftUI for iOS (for image capture / photo upload)

Challenges we ran into

  • Data scalability: Handling plants with various stages of diseases and a large variety of symptoms required training with a large dataset
  • Mobile app efficiency: Ensuring the AI model could run smoothly on mobile devices
  • LLM Response Management: Fine-tuning the prompts for the language models to provide detailed responses without exceeding token limits

Accomplishments that we're proud of

  • Successfully training our CNN model on Google Colab using transfer learning with InceptionV3 as the base model.
  • Creating a user-friendly mobile interface for easy accessibility by farmers
  • Implementing a Flask API to connect our technologies together

What we learned

  • How to optimize LLM prompts for best performance
  • How to use API keys to transfer data from iOS to external Python code and back
  • How to train CNNs effectively

What's next for Tensor Crop Tracker (TCT) AI

  • Location-based monitoring of diseased plants to monitor their status
  • Citing reliable Internet sources to promote more sustainable farming and agriculture research
  • Harnessing IoT (e.g. drones) to automate disease detection.

Built With

Share this project:

Updates