Processing happens at the trap, not in the cloud
Rural polyhouse clusters don't always have strong connectivity. So the heaviest work — looking at the image and deciding what pest it is — happens on the sensor itself. The network only carries the result.
Why we process on-device
Far less data to send
A result like "whitefly, count 14" is a few bytes. A photo is hundreds of kilobytes — and most polyhouse clusters can't rely on sending that reliably.
Faster alerts
Detection happens the moment the photo is taken. There's no round trip to a server before you know something is wrong.
Keeps battery low
Nodes wake, capture, process, and sleep — instead of streaming constantly, which would drain a solar-charged battery fast.
The stack, end to end
Built to survive a polyhouse season
Designed around the conditions that actually break field electronics in India: heat, humidity, dust, and inconsistent sun hours during monsoon.
| Power | Solar panel + LiFePO4 battery, multi-day reserve |
|---|---|
| Connectivity | LoRaWAN, one gateway covers a multi-acre cluster |
| Enclosure rating | IP65, dust and splash resistant |
| Capture interval | Configurable, default every 20 minutes |
| Coverage | One node per polyhouse zone or bay |
| Alert channel | SMS and WhatsApp, local language |
From pest counts to pest and disease, in one alert
The same camera trap that spots and counts pests is also where we're adding disease detection — reading early leaf symptoms like blight, powdery mildew, and leaf spot from the same images, so one node and one app cover both.
No separate hardware, no second app to check — pest and disease risk show up together on the same dashboard.
Want the technical details?
Happy to walk through the model, the hardware, or a pilot deployment plan.
Get in touch