AI-powered smartphone microscope platform for accurate cucurbit disease diagnostics
An AI-enabled smartphone microscope integrated with lab diagnostics and machine learning to accurately differentiate five destructive cucurbits. The platform achieves 100% precision and 93%+ recall rates, addressing limitations of AI-only plant disease diagnostics by combining ground truthing with lab confirmation.
Overview
This solution combines a smartphone-attachable microscope with laboratory diagnostics and machine learning to accurately identify and differentiate five destructive cucurbit diseases: downy mildew, gummy stem blight, anthracnose, cucurbit leaf crumple, and silver leaf. These diseases cause significant economic losses in the U.S. and internationally. By integrating ground-truthed field images with validated lab confirmation methods, the platform overcomes the accuracy limitations that have stalled AI-based plant disease diagnostics, particularly for diseases with closely resembling symptoms.
Technical specifications
Key features:
- 60X smartphone-attachable microscope (patent pending) for capturing leaf lesion images
- Integrated web app for image capture and data management
- Machine learning models developed using TensorFlow, Create ML, and AutoML
- Lab confirmation workflow including culturing, microscopy, antibody-based, and DNA/RNA-based diagnostics
- Confusion matrix analysis to improve AI model training, testing, and validation
- Disease-specific confirmation: fruiting body observation and culturing for gummy stem blight and anthracnose; spore microscopy for downy mildew; PCR with specific primers for cucurbit leaf crumple; characteristic leaf pattern recognition for silver leaf
- 100% precision rate and 93%+ recall rate across all five diseases
Image capture parameters include:
- Various lesion stages (young, moderately-aged, and old)
- Symptom location (leaf, stem, fruit)
- Multiple cucurbit species (watermelon, squash, cucumber)
- Variable lighting conditions (flash/no flash, sunny/cloudy)
- Leaf wetness conditions (wet, dry)
Technology readiness level
The platform has been validated with over 100 images captured from commercial cucurbit production fields in Florida. A beta version of the AI platform is planned for development, with future validation targeting over 1,000 naturally occurring leaf lesion images across diverse conditions. The smartphone microscope technology is patent pending.
About University of Florida
The University of Florida is a comprehensive public research university with a broad academic and research portfolio and a statewide presence. Industry collaborates through co-located labs and shared core facilities and through an integrated academic health system that accelerates clinical translation. A statewide extension network and multiple research and education sites connect companies with field-scale testing and rapid deployment, while incubators and an adjacent innovation district provide pathways from lab to market. Research is supported by competitive funding from major federal agencies such as NIH, NSF, USDA, and DOE. A dedicated technology transfer office supports IP, licensing, and startup formation.
delivered to your inbox.