requirements.txt

# GAZE2CodeToolkit — Python dependencies
# Tested on Python 3.10+.
# Pin only where strictly needed; otherwise let pip resolve the latest
# compatible version. System dependency: Tesseract OCR binary
# (see README.md "System dependencies").

# --- Core scientific stack ---
numpy>=1.23
pandas>=2.0
scipy>=1.10

# --- Image processing and OCR ---
opencv-python>=4.8
Pillow>=10.0
pytesseract>=0.3.10        # Python bindings; requires `tesseract` binary

# --- Plotting ---
matplotlib>=3.7
seaborn>=0.13

# --- Progress bars ---
tqdm>=4.66

# --- OCR evaluation (cli/evaluate_ocr.py) ---
rapidfuzz>=3.5
python-Levenshtein>=0.23
scikit-learn>=1.3          # ROC curve in cli/evaluate_ocr.py and CV/scaling in g2c/classification/

# --- Expertise classification (cli/classify_expertise.py, g2c/classification/) ---
xgboost>=2.0               # Locked baseline classifier; pin if reproducing thesis numbers

# --- Web UI (app.py / webapp/) ---
streamlit>=1.30            # Headless browser UI; optional if you only use the CLIs