Real-time animal detection running on a $99 board in the dark

Nighttime feral-animal detection deployed end-to-end on a Jetson Nano, triggering non-lethal deterrence — won a competitive university research grant.

Role

Project lead — proposal, build, field deployment

Goal

Autonomous nighttime detection + non-lethal deterrence

Key outcome

Sustained 24 FPS at 720p on-device; grant awarded

Stack

Python · YOLOv8 · Jetson Nano · OpenCV

24 FPS

sustained at 720p — real-time inference on a Jetson Nano

2 yrs

from proposal to field deployment — full project lifecycle

competitive university research grant won

$ cat problem.md

No cloud, no operator, no daylight

Feral animals do damage at night, in places without connectivity. The system had to see in the dark, decide on-device, and act — within the compute and power budget of an embedded board.

$ ./approach.sh

01Deployed YOLOv8 on a Jetson Nano with an IMX219 night-vision camera — optimized to a sustained 24 FPS at 720p.

02Wired detections to trigger non-lethal deterrence autonomously — a full sense-decide-act loop.

03Ran the project end-to-end over two years: proposal → competitive research grant → build → field deployment.

$ cat results.json

Real-time on-device inference: 24 FPS at 720p, sustained, at night
Autonomous detection → deterrence loop with no operator
Won a competitive university research grant on the strength of the proposal and prototype

$ grep -r skills

edge deploymentYOLOv8Jetson Nanoreal-time optimizationOpenCVembedded systemsgrant writingfield deployment