MatchResult

spatialhub.structures.MatchResult represents 2D keypoint correspondences and match confidence scores produced by feature matching adapters such as EfficientLoFTRAdapter.

from spatialhub.structures import MatchResult

---

Fields

Field	Type	Shape	Description
image_a	str | Path | np.ndarray	(H, W, C) or path	First input image — path or in-memory array.
image_b	str | Path | np.ndarray	(H, W, C) or path	Second input image — path or in-memory array.
keypoints_a	np.ndarray	(N, 2) float32	Matched keypoint pixel coordinates [x, y] in image_a.
keypoints_b	np.ndarray	(N, 2) float32	Matched keypoint pixel coordinates [x, y] in image_b.
confidence	np.ndarray	(N,) float32	Per-match confidence scores in [0.0, 1.0].

---

Methods

visualize

Renders a side-by-side keypoint correspondence canvas. Calls visualize_matches internally.

vis = result.visualize(conf_thresh=0.5, max_side=800, top_k=50, save_path="matches.png")

Parameters

Parameter	Type	Default	Description
conf_thresh	float	0.5	Minimum confidence score threshold to draw a match.
max_side	int	800	Caps the longest image dimension in the output visualization canvas.
top_k	int | None	None	Limit display to top \(k\) matches by confidence score.
save_path	str | Path | None	None	Optional disk path to write the annotated image.

Return Value

• np.ndarray: uint8 BGR visualization canvas of shape (H, W_a + W_b, 3).