Warbler‑YOLO: Segmentation‑driven fine‑grained bird classification (YOLO11)

End‑to‑end pipeline that takes unannotated images, segments birds with YOLO11‑Seg, auto‑crops foregrounds, then trains a YOLO11‑Cls fine‑grained classifier.

Works on Mac (CPU or Apple M‑series with MPS), Linux, and Windows.
Dataset layout: data/raw/<class-id>/*.png|jpg (e.g., 00000, 00001, …).

---

Quick start

# 1) Create env (example with conda)
conda create -n warblers python=3.11 -y
conda activate warblers

# 2) Install deps
pip install --upgrade pip
pip install -r requirements.txt

# 3) Put your data like:
# data/raw/00000/*.png
# data/raw/00001/*.png
# ...
# data/raw/00009/*.png

# 4) Prepare (segment + crop → train/val/test)
python -m warbler_yolo prepare --raw data/raw --work runs/warbler --imgsz 640

# 5) Train classifier
python -m warbler_yolo train   --work runs/warbler --epochs 50

# 6) Evaluate on test split
python -m warbler_yolo eval    --work runs/warbler

source /Users/alialfatemi/venvs/warblers/bin/activate