🦝 Backyard Raccoon Deterrent

Fine-tuned YOLOv8n raccoon detector (v1.4) — the eyes of a real Ring-camera deterrent. Trained on 560+ hand-labeled night-vision frames of raccoons raiding my yard, including trajectory frames pulled from real motion events (P 93.5% · R 85.9% · mAP50 92.8% on a held-out val split, ~24 ms inference). Runs fully offline. Upload a frame or click an example.

Built for the Gradio Build Small hackathon (Backyard AI track). The deployed system pairs this model with audio + smart-light deterrents on a Raspberry Pi — fully offline, no cloud APIs. Source on GitHub.