LeafGuard
About the project
The problem
Farmers often work where there is no signal, and an expert can be days away. Telling early blight from late blight or rust by eye is hard. Waiting costs the crop, and guessing wrong means spraying chemicals that are wasted or harmful.
Our solution
LeafGuard runs entirely in the phone’s browser. The farmer scans a leaf, and the app shows the likely condition, how severe it looks, and organic and chemical treatment steps with doses. Everything works offline, and the advice is available in Hindi and English, spoken aloud or typed.
What it does
Leaf scan: on-device diagnosis with severity and treatment steps, covering 12 crops including wheat, rice, cotton, tomato and potato.
Sahayak assistant: a bilingual chat (English and Hindi) with voice input and output. Common questions are answered offline from built-in knowledge. Open questions go to Gemma 4 through the Gemini API when the phone is online, and those replies are tagged with a Gemma badge.
Soil meter: sliders for moisture, pH and temperature, with healthy ranges for each crop and plain-language advice.
Field logbook: scans saved on the device, with a sync state for when a connection returns.
Outbreak map: a community view of nearby reports (sample data in this demo).
Open science hub: an in-browser speed benchmark and model-card export.
Built with
AI tools
Prize Categories Entered
- Event prizeSubmitted
Hacktoberfest Hack Day Winners
- Event prizeSubmitted
Best Beginner Hack
- Event prizeSubmitted
ElevenLabs Track Prize
- Event prizeSubmitted
169pi Track Prize