Spore Scout

Scanner
Community Page
Scans History
About Page

Inspiration

Annually, a considerable quantity of agricultural yield is discarded due to various diseases. This issue disproportionately affects farmers with low-to-medium incomes, as they lack access to costly machinery and equipment necessary for accurate disease diagnosis. The emergence of advanced deep-learning models presents an opportunity to provide assistance to these vulnerable farmers.

What it does

Spore Scout is a mobile app that lets users snap or upload a photo of a plant leaf. Using an on-device AI model, the app classifies the leaf as healthy or having multiple diseases. The output includes the detected condition, a confidence score, and suggestions for the next steps.

How we built it

Frontend/UI: Built in Flutter, providing a clean, simple interface for uploading or capturing leaf images.

Model training: We used a plant disease dataset and split it into training/validation/testing sets. We fine-tuned an ResNet model in TensorFlow/Keras to distinguish between target classes.

Deployment: The trained model was exported to TensorFlow Lite and integrated into the Flutter app using the tflite_flutter plugin. This allows real-time, offline predictions directly on a mobile device.

Challenges we ran into

Implementing the in-app camera feature in Flutter was more complex than expected. Handling permissions, ensuring consistent image quality across devices, and integrating the captured image with the ML pipeline was challenging. Additionally, building both the ML pipeline and the Flutter app in a limited hackathon window pushed us to prioritize features.

Accomplishments that we're proud of

Successfully trained and deployed an AI plant disease classifier on a mobile device.

What we learned

That simple, well-designed solutions can empower users without requiring them to understand the complexity of AI.

What's next for Spore Scout

Expand dataset to cover more crops and diseases. Add multi-label predictions for leaves showing overlapping symptoms. Provide treatment recommendations and integrate with agricultural knowledge bases. Collect user feedback and photos to continuously improve the model.