I turned my security cameras into an automatic bird identification system

548 points · 144 comments on HN · read original →

Points and comments are a snapshot, not live.

A DIY project repurposes security cameras to automatically identify birds by song using BirdNet-Go.

The author used BirdNet-Go, running in Docker, to listen to RTSP audio streams from three existing security cameras. The system runs 24/7 locally, identifies birds and bats in real time using AI models, and supports alert rules for specific species. Over 12 months, it logged 418,726 detections of 271 unique species in Southern California, with an average confidence of 60.9%. The system integrates with Home Assistant and BirdWeather, and even detected a neighbor's fart.

What commenters are saying

Commenters were enthusiastic, sharing similar DIY setups and discussing integration with Unifi cameras, e-ink displays, and the BirdWeather platform. Several contributors to BirdNet-Go shared their own custom frontends and a 24/7 bird-and-weather TV channel. Some noted the microphones on security cameras can be low-quality, and a few users recommended the Merlin Bird ID app as a simpler alternative. The thread also included a brief discussion of the ethical and practical concerns around repurposing Flock security cameras.