- IoT
- Computer Vision
- Research
SATRIA
Telling a rat from a shadow at two in the morning, on solar power.
- Role
- Year
- Stack
Overview
A solar-powered camera trap for rice fields: an infrared sensor wakes the camera, a YOLO model decides whether what walked past was a rat, and the result lands on a dashboard the farmer can read from a phone.
The problem
Rodent damage gets noticed after the harvest is already short. Counting the problem means being in the field at night, repeatedly, which nobody has time to do.
What I built
- A field unit built on ESP32 with PIR and infrared capture, running on solar
- On-device detection tuned for night, glare, and motion blur
- Real-time sensor monitoring and remote device control from the dashboard
- Interactive data visualisation with automatic alerts when activity spikes
Outcome
The hard part was never the model — it was keeping recall high on grainy infrared frames while the whole unit stays inside a solar power budget. That trade-off is still where the work is.
Next project
Teleradiologi AI