Line detection, description and matching

LIMAP provides modular interfaces for line detection, description and matching under limap.image.line.

Minimal example on line detection and description

Detection and description operate directly on an image file path. Here is a minimal example running DeepLSD detection and SOLD2 description on an image example.png:

from pathlib import Path
import limap.image.line as line2d

image_path = Path("example.png")
detector = line2d.get_detector("deeplsd", line2d.DetectorOptions())
segs = detector.detect(image_path)          # (N, 5): x1, y1, x2, y2, score
extractor = line2d.get_extractor("sold2", line2d.ExtractorOptions())
desc = extractor.extract(image_path, segs)  # descriptors for the detected segments

Minimal example on line matching

The matcher type must be compatible with the extractor. Here is a minimal example running the SOLD2 matcher on two sets of descriptors:

import limap.image.line as line2d

# desc1, desc2: descriptors extracted from two images (see above)
extractor = line2d.get_extractor("sold2", line2d.ExtractorOptions())
matcher = line2d.get_matcher("sold2", line2d.MatcherOptions(), extractor)
matches = matcher.match_pair(desc1, desc2)

Visualization

Here is an example on visualizing the detected segments:

import cv2
import limap.visualize

image = cv2.imread("example.png")
# segs is (N, 5); reshape the endpoints to (2, 2) per line
lines = [seg[:4].reshape(2, 2) for seg in segs]
image = limap.visualize.draw_2d_lines(image, lines, (0, 255, 0))
cv2.imshow("detections", image)
cv2.waitKey(0)

Multiple images

To run line detection, description and matching over many images at once, use the batch helpers in limap.image.line:

The output detections, descriptions and matches are saved into the corresponding output folders.