KeypointDetector (kp_detection) is a Python package that provides a unified interface for keypoint detection using various OpenCV-backed algorithms.
For more details, see src/kp_detection/README.md.

From the package root (the directory containing pyproject.toml):

pip install .

For development, install in editable mode so changes to the source take effect immediately:

pip install -e .

Dependencies (numpy, opencv-contrib-python) are installed automatically.
To install only the dependencies without the package, use:

pip install -r requirements.txt

After installing the package, import it from any directory:

import cv2

from kp_detection import KPDetectionMethod, KPDetectionResult

image = cv2.imread("image.png", cv2.IMREAD_GRAYSCALE)

method = KPDetectionMethod.SIFT
params = method.parameter_class(method=method)
detector = method.detector_class(params=params)
result: KPDetectionResult = detector.detect(image)
# Use result according to KPDetectionResult / ArrayKPDetectionResult (see package README).

KPDetectionMethod maps each variant to a parameter dataclass and a detector class via parameter_class and detector_class (see method.py). For Harris or Shi–Tomasi, parameter_class resolves to HarrisParameters or ShiTomashiParameters automatically when you pass the matching method.