Birdjournal: Identify Birds by Sound, Offline, with Open-Source AI

Описание и применение

Birdjournal позволяет любителям птиц идентифицировать вид по звуку без подключения к интернету, используя BirdNET и локальную LLM.

Задача

Beginner birdwatchers need a way to identify bird species from audio recordings without internet or cloud services.

Результат

On a 46‑second recording of a Coppersmith Barbet, BirdNET detected the species at 100% confidence in all 16 three‑second windows, and the LLM produced a concise journal entry in English.

Рабочий процесс

• Record audio (.wav/.mp3) on a walk
• Run python main.py <audio> --place "<city>, <state>" --lat <lat> --lon <lon> --out <output>
• BirdNET identifies species and applies a geo filter
• A local LLM (default Qwen2.5 3B via Ollama) generates a short field‑journal entry
• Output is written to the specified markdown file

Доказательства

BirdNET detected the barbet at 100% confidence in all 16 of its 3-second windows, with the location filter set to Sikar.

Источник

The idea is that the screen is the shortest part of the experience. You go outside, record, and listen. The tool does its work afterwards.

Источник

My first prompt let the small model invent facts that weren't in the detections, such as calling the barbet "colorful and melodious".

Источник