KasinathC A
all projects

Open source / edge AI

Forest Ranger

ESP32 acoustic sensing node that runs a quantised CNN locally to detect chainsaw activity and send compact alerts.

GitHubResearch manuscript
ModelQuantised CNN
Inference~1 ms model inference
Footprint~40.3 kB Flash / 12.8 kB peak RAM

overview

An edge-AI prototype for illegal-logging detection. An INMP441 microphone feeds 16 kHz audio to an ESP32; MFE features are computed locally and classified by an int8 CNN exported from Edge Impulse.

what it covers

  • Local inference avoids continuous audio streaming and reduces bandwidth.
  • Prototype alerts are sent as lightweight telemetry after positive detection.
  • Field design considers solar power and long-range LoRa-class telemetry.