Inspiration
Many small-scale farmers and home gardeners struggle to identify plant diseases early. We wanted to build a solution that reduces crop losses and helps them take proactive measures with minimal effort.
What it does
Our “Plant Disease Detector” analyzes images of plants and instantly identifies the likely disease. It then provides customized treatment suggestions based on factors like humidity, temperature, and soil pH.
How we built it
We trained a YOLO model on a large dataset of leaf images labeled with common diseases. A JSON file maps each disease to recommended treatments, adjusting for environmental data. We integrated Python scripts to handle image inputs, run inferences, and deliver a concise report with annotated images.
Challenges we ran into
-Ensuring high accuracy for diverse diseases and similar symptoms. -Handling noisy or low-confidence detections when leaves were partially visible. -Fine-tuning the model to avoid overfitting and to handle images that differ from the training set.
Accomplishments that we're proud of
Achieving around 98% precision and recall on our test set. Delivering real-time disease detection and annotated images within seconds. Providing targeted treatment advice that factors in current weather and soil conditions.
What we learned
The importance of robust data augmentation to handle varied real-world images. How to map model outputs to specific treatments via a flexible JSON structure.
What's next for Plant Disease Detector
Expanding our model to cover more diseases and additional crops. Improve model for more than just leaf photos. Launching a mobile-friendly interface for instant on-the-go detection.
Log in or sign up for Devpost to join the conversation.