chess-corners 0.4.0

Python bindings for the fast ChESS chessboard corner detector (Rust backend)

chess_corners (Python)

Python bindings for the chess-corners detector.

Quick start

import numpy as np
import chess_corners

img = np.zeros((128, 128), dtype=np.uint8)

cfg = chess_corners.ChessConfig()
cfg.threshold_rel = 0.2
cfg.min_cluster_size = 1

corners = chess_corners.find_chess_corners(img, cfg)
print(corners.shape, corners.dtype)

You can also configure the detector via nested config objects:

cfg = chess_corners.ChessConfig()
cfg.params.threshold_rel = 0.2
cfg.multiscale.pyramid.num_levels = 2

What find_chess_corners returns

find_chess_corners(image, cfg=None) returns a NumPy float32 array of shape (N, 4) with columns:

1. x – subpixel x coordinate (pixels, image space)
2. y – subpixel y coordinate (pixels, image space)
3. response – ChESS response strength at the corner
4. orientation – local grid axis orientation in radians, in [0, pi)

The rows are sorted deterministically by the binding: response (descending), then x, then y.

Input requirements:

• image must be a 2D uint8 NumPy array with shape (H, W)
• it must be C-contiguous (non-contiguous arrays raise ValueError)

ChessConfig parameters

ChessConfig mirrors the most important Rust settings. Defaults match the Rust ChessConfig::default().

Response / detector parameters (cfg.*):

• use_radius10 (bool, default False)
  • Use the larger ring radius (r=10) instead of r=5 for response computation.
• descriptor_use_radius10 (Optional[bool], default None)
  • Override the descriptor sampling ring radius; when None, uses use_radius10.
• threshold_rel (float, default 0.2)
  • Relative threshold as a fraction of the max response.
• threshold_abs (Optional[float], default None)
  • Absolute threshold; when set, it overrides threshold_rel.
• nms_radius (int, default 2)
  • Non-maximum suppression radius in pixels.
• min_cluster_size (int, default 2)
  • Minimum count of positive neighbors in the NMS window to accept a corner.

Multiscale parameters (cfg.*):

• pyramid_num_levels (int, default 3)
  • Number of pyramid levels (including base). Set to 1 for single-scale.
• pyramid_min_size (int, default 128)
  • Minimum dimension to keep building the pyramid.
• refinement_radius (int, default 3)
  • Coarse-level ROI radius used for coarse-to-fine refinement.
• merge_radius (float, default 3.0)
  • Merge near-duplicate refined corners within this radius (pixels).

Full configuration structs

The Python bindings expose all configuration structs directly. You can use the nested API for clarity:

• ChessConfig.params → ChessParams
• ChessConfig.multiscale → CoarseToFineParams
• CoarseToFineParams.pyramid → PyramidParams

All fields listed above are available in the nested structs as well.

Classical refiners

Select the classic refiner via ChessParams.refiner:

cfg = chess_corners.ChessConfig()
cfg.params.refiner = chess_corners.RefinerKind.forstner(
    chess_corners.ForstnerConfig()
)

Available configs:

• CenterOfMassConfig
• ForstnerConfig
• SaddlePointConfig

ML refiner

If the bindings are built with the ml-refiner feature, an ML-backed refiner is available via find_chess_corners_with_ml:

cfg = chess_corners.ChessConfig()
corners = chess_corners.find_chess_corners_with_ml(img, cfg)

The ML refiner uses built-in defaults and ignores the model’s confidence output in the current version.

Development

maturin develop -m crates/chess-corners-py/pyproject.toml