vismatch.im_models.keypt2subpx

Classes

Keypt2SubpxMatcher([device, detector_name])
SuperPointDense([device])

class vismatch.im_models.keypt2subpx.Keypt2SubpxMatcher(device='cpu', detector_name=None, **kwargs)

Bases: BaseMatcher

Parameters:

• device (str)

• detector_name (str | None)

THIRDPARTY_NAMES = {'aliked-lightglue': 'aliked', 'dedode': 'dedode', 'superpoint-lightglue': 'splg', 'xfeat': 'xfeat', 'xfeat-lightglue': 'xfeat'}

get_match_idxs(mkpts, kpts)

Parameters:

• mkpts (ndarray | Tensor)

• kpts (ndarray | Tensor)

Return type:

ndarray

get_scoremap(img=None, idx=None)

class vismatch.im_models.keypt2subpx.SuperPointDense(device='cpu', **kwargs)

Bases: BaseMatcher

Parameters:

device (str)

modelconf = {'allow_no_extract': True, 'extractor': {'detection_threshold': 0.0, 'force_num_keypoints': False, 'max_num_keypoints': 2048, 'name': 'superpoint_densescore', 'nms_radius': 3, 'remove_borders': 3, 'trainable': False}, 'ground_truth': {'name': 'matchers.depth_matcher', 'th_epi': 5, 'th_negative': 5, 'th_positive': 3}, 'matcher': {'depth_confidence': -1, 'filter_threshold': 0.1, 'name': 'matchers.lightglue_wrapper', 'trainable': False, 'weights': 'superpoint', 'width_confidence': -1}, 'name': 'two_view_pipeline'}

preprocess(img)

get_scoremap(idx)