Overview
PlantDx allows users to upload leaf images, select the appropriate classification model, and receive multi-class disease predictions with confidence scores, along with downloadable diagnostic reports.
The problem
Various deep learning and transformer-based models have been developed for plant leaf disease classification through research studies and individual machine learning implementations. However, these solutions are often limited to standalone models, research experiments, or notebooks, making them difficult for users to access and use in practice. There is a lack of a unified platform that provides access to multiple leaf disease classification models across different fruit plants through a single application.
The solution
I developed a full-stack platform that integrates four transformer-based models with an interactive web interface for image uploads, model selection, disease classification, confidence scoring, authentication, and report generation.
Approach
The platform uses four transformer-based models for classifying diseases across apple, mango, cherry, and grape plants. Users upload a leaf image, select the appropriate model, and receive a multi-class prediction indicating whether the leaf is healthy or diseased.
Key features
- Upload fruit plant leaf images
- Select the appropriate classification model
- Multi-class disease predictions
- Healthy vs. diseased classification
- Confidence scores for predictions
- JWT-based user authentication
- Downloadable diagnostic reports
What I learned
Building PlantDx strengthened my experience with transformer-based computer vision models and integrating machine learning predictions into a complete full-stack application.