Logo Lanfrica
  • Accueil
  • Atlas
  • Analyses
  • Documentation
  • Sign in

© 2026 Lanfrica. Tous droits réservés. Tous les droits d'auteur des ressources affichées sur le site Web Lanfrica appartiennent aux détenteurs de droits d'auteur d'origine, sauf indication contraire explicite.

Hailion/TomatoDx---Offline-Tomato-Disease-Diagnosis-App

Domaine:

agriculture

Type de record:

softwaremodel
Créateur:
Hai
Hôte:
A React Native TypeScript mobile application for offline tomato disease diagnosis using TensorFlow Lite. Designed for low-literacy farmers with Amharic localization and large, accessible UI elements. Built with Expo managed workflow for easy development and deployment. # TomatoDx **Offline tomato disease diagnosis for smallholder farmers.** TomatoDx is a cross-platform mobile app that helps farmers and gardeners detect tomato plant diseases quickly and accurately. Using on-device machine learning, the app analyzes leaf photos locally, no internet connection required, and returns a diagnosis with treatment guidance in the user's preferred language. **Live demo:** tomatodx.netlify.app --- ## Screenshots --- ## Features - **Instant leaf scanning** — Capture a photo with the camera or upload from the gallery. - **On-device AI inference** — TensorFlow.js runs entirely on the phone; images never leave the device unless you choose to share them. - **10 disease classes** — Detects common tomato diseases plus healthy-leaf classification. - **Actionable results** — View confidence scores, severity levels, symptoms, and treatment advice for each diagnosis. - **Scan history & insights** — Review past scans, track trends over time, and monitor crop health from the dashboard. - **Multilingual support** — English, Amharic (አማርኛ), and Afaan Oromoo. - **Ethiopian calendar** — Dates displayed using the Ethiopian calendar where applicable. - **Privacy-first design** — All processing happens locally; optional sharing for research is entirely user-controlled. - **Reminders** — Scheduled notifications to encourage regular crop checkups. - **Dark & light themes** — Automatic theme switching based on system preference. --- ## Model Evaluation The classifier was evaluated on a held-out test set of 15,253 images, separate from the training data, across all 10 disease classes plus an out-of-distribution "Unknown" class used to reject non-leaf or unrecognized inputs. **Overall Accuracy:** 91.65% **Macro F1 Score:** 0.8892 **Weighted F1 Score:** 0.9147 | Class | Precision | Recall | F1 Score | | ----------------------------- | --------- | ------ | -------- | | Tomato Yellow Leaf Curl Virus | 0.9958 | 0.9714 | 0. …

Visit

github.com

Tasks

computer visionimage classification

Languages

AmharicOromo, Borana-Arsi-GujiOromo, EasternOromo, West Central

Licenses

MIT