An early-warning system for cabbage disease. A small box in the field, plus free satellite and weather data, learns what a healthy field looks like — and flags the drift toward disease 12–72 hours before it shows on the leaf.
DashboardWhat is AgriSync
Most tools name a disease only once it's visible on a leaf — when the spray window is already half closed. AgriSync learns what one healthy cabbage field looks like and warns as it starts to drift, 12–72 hours before the eye can catch it.
A smallholder can lose most of a crop to disease in a single week. AgriSync buys back the hours that decide the season.
How it works
One box in the field, free satellite and weather data, and a small computer at home — that's the whole system. No cloud bill, no farmer hardware to buy.
Leaf wetness, temperature, humidity and rain — measured right at the leaf, the exact conditions a fungus needs to take hold.
Sentinel-2 reads crop stress from orbit; Sentinel-1 radar sees through monsoon cloud. Both free, at 10-metre resolution.
Decades of hourly history to learn from (ERA5-Land, NASA POWER), plus a daily forecast (Open-Meteo) that drives the warning.
A small computer at home pulls every source together and runs the models each night. No cloud, no laptop, no field servers.
The output is a rising risk curve for the next one to three days — an honest window, not a false-precise single hour.
Every real outbreak is logged against the conditions that preceded it — the first local record of this crop's disease.
Intelligence
Known plant-science rules give a dependable floor from day one. A learned "healthy baseline" catches what the rules miss — and every answer comes as a calibrated range, not a fake-precise number.
Published temperature, humidity and leaf-wetness thresholds score disease pressure live. Works on day one, no training data — a ~60–70% floor.
It models a healthy field across sensor and satellite signals, then flags the departure — so it needs no library of disease-labelled examples.
A drift alone could be drought or hunger. It's confirmed only when the weather sits in the infection window and the satellite bands agree.
Borrowing conformal prediction from ML safety, the answer is a range with a stated error rate — never a single, falsely precise number.
Why it matters
Every rule, threshold and healthy-baseline is tuned to one real cabbage field in Hapur, UP — not a generic, all-crop model.
Per-plot rigs cost a fortune and target vineyards; satellite platforms sell generic data to corporates. No deployed product combines pre-symptom prediction, one crop and cheap shared hardware for a smallholder.
The satellites, the weather and the science are free or open. The farmer buys no per-plot hardware — one cheap shared box can serve a whole cluster of fields.
Built solo from one family's cabbage field in Hapur, on minimum resources — the first dataset of its kind, made by hand.
Under the hood