Edge AI Pest Detection Camera

A weather-sealed, solar-powered camera node built around an ESP32 and a Coral TPU running a quantized TensorFlow Lite pest detector. The device captures and classifies insect activity entirely on-device, buffers detections, and sends compact alerts over LoRaWAN — designed for fields with no cellular coverage and a strict power budget.

Pest outbreaks are cheapest to control when caught early, but fields rarely have the connectivity or power for cloud-based monitoring. Growers need a node that detects pests locally and alerts them without continuous uplink or mains power.

A Coral TPU runs a quantized TensorFlow Lite pest detector on frames captured by an ESP32-driven camera, classifying insect presence on-device in milliseconds. Detections are buffered and sent as compact LoRaWAN alerts. Solar charging and aggressive duty-cycling keep the node running unattended for extended periods.

Planned — targets: reliable on-device pest detection, multi-week battery life on solar, and useful early-warning alerts delivered over long-range LoRaWAN.

1. Phase 1

   Detector

   Train and quantize a TensorFlow Lite pest-detection model for the Coral TPU.

2. Phase 2

   Capture node

   Build the ESP32 camera node with on-device inference and detection buffering.

3. Phase 3

   Power & uplink

   Add solar charging, duty-cycling, and LoRaWAN alert delivery.

4. Phase 4

   Field validation

   Measure detection accuracy, battery life, and uplink range in the field.